filter by: Publication Year
(Descending) Articles
Digital Signal Processing: A Review Journal (10954333) 164
Multi-frame image super-resolution represents an efficacious albeit expensive and resource-intensive technique for image reconstruction, necessitating substantial memory allocation for data storage. To mitigate the computational burden inherent in multi-frame image super-resolution algorithms, a strategic approach involves curtailing the processing load by disregarding redundant frames. In this study, we introduce a novel frame selection algorithm tailored to identify an optimal minimum number of frames. This approach ensures the fidelity of the reconstructed high-resolution (HR) image while significantly alleviating the procedural complexity and memory demands of the super-resolution process. The frame selection methodology we propose is founded upon multi-channel sampling, reference frame selection, and the maximization of the lower bound on the signal-to-noise ratio. Specifically, our approach is operationalized through two optimization algorithms based on priority search. The initial algorithm identifies cases with maximum non-empty channels by exploring the predefined domain encompassing all feasible desired positions. In the subsequent algorithm, the process entails identifying, for any channel within any discovered case, a frame associated with the minimum translation function model noise. Subsequently, the total noise of each case is computed. We ascertain the optimal case along with a collection of frames that correspond to the minimum total noise. Experimental findings highlight the efficacy of our proposed method in mitigating super-resolution complexity while achieving high-fidelity HR images that closely match or surpass those generated from complete frame sets. Comparative analysis against established super-resolution (SR) algorithms demonstrates the remarkable speed and minimal computational overhead of our proposed approach, rendering it exceptionally efficient with negligible runtime. © 2025 Elsevier Inc.
PeerJ Computer Science (23765992) 10
In this article, a novel method for removing atmospheric turbulence from a sequence of turbulent images and restoring a high-quality image is presented. Turbulence is modeled using two factors: the geometric transformation of pixel locations represents the distortion, and the varying pixel brightness represents spatiotemporal varying blur. The main framework of the proposed method involves the utilization of low-rank matrix factorization, which achieves the modeling of both the geometric transformation of pixels and the spatiotemporal varying blur through an iterative process. In the proposed method, the initial step involves the selection of a subset of images using the random sample consensus method. Subsequently, estimation of the mixture of Gaussian noise parameters takes place. Following this, a window is chosen around each pixel based on the entropy of the surrounding region. Within this window, the transformation matrix is locally estimated. Lastly, by considering both the noise and the estimated geometric transformations of the selected images, an estimation of a low-rank matrix is conducted. This estimation process leads to the production of a turbulence-free image. The experimental results were obtained from both real and simulated datasets. These results demonstrated the efficacy of the proposed method in mitigating substantial geometrical distortions. Furthermore, the method showcased the ability to improve spatiotemporal varying blur and effectively restore the details present in the original image. © Copyright 2024 Jafaei et al
IET Computer Vision (17519640) 18(2)pp. 191-209
The position of vehicles is determined using an algorithm that includes two stages of detection and prediction. The more the number of frames in which the detection network is used, the more accurate the detector is, and the more the prediction network is used, the algorithm is faster. Therefore, the algorithm is very flexible to achieve the required accuracy and speed. YOLO's base detection network is designed to be robust against vehicle scale changes. Also, feature maps are produced in the detector network, which contribute greatly to increasing the accuracy of the detector. In these maps, using differential images and a u-net-based module, image segmentation has been done into two classes: vehicle and background. To increase the accuracy of the recursive predictive network, vehicle manoeuvres are classified. For this purpose, the spatial and temporal information of the vehicles are considered simultaneously. This classifier is much more effective than classifiers that consider spatial and temporal information separately. The Highway and UA-DETRAC datasets demonstrate the performance of the proposed algorithm in urban traffic monitoring systems. © 2023 The Authors. IET Computer Vision published by John Wiley & Sons Ltd on behalf of The Institution of Engineering and Technology.
Visual Computer (14322315) 40(10)pp. 6825-6841
An anomaly is a pattern, behavior, or event that does not frequently happen in an environment. Video anomaly detection has always been a challenging task. Home security, public area monitoring, and quality control in production lines are only a few applications of video anomaly detection. The spatio-temporal nature of the videos, the lack of an exact definition for anomalies, and the inefficiencies of feature extraction for videos are examples of the challenges that researchers face in video anomaly detection. To find a solution to these challenges, we propose a method that uses parallel deep structures to extract informative features from the videos. The method consists of different units including an attention unit, frame sampling units, spatial and temporal feature extractors, and thresholding. Using these units, we propose a video anomaly detection that aggregates the results of four parallel structures. Aggregating the results brings generality and flexibility to the algorithm. The proposed method achieves satisfying results for four popular video anomaly detection benchmarks. © The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2024.
Computers and Electrical Engineering (00457906) 120
Video anomaly detection is the identification of outliers deviating from the norm within a series of videos. The spatio-temporal dependencies and unstructured nature of videos make video anomaly detection complicated. Many existing methods cannot detect anomalies accurately because they are unable to learn from the learning data effectively and capture dependencies between distant frames. To this end, we propose a model that uses a pre-trained vision transformer and an ensemble of deep convolutional auto-encoders to capture dependencies between distant frames. Moreover, AdaBoost training is used to ensure the model learns every sample in the data properly. To evaluate the method, we conducted experiments on four publicly available video anomaly detection datasets, namely the CUHK Avenue dataset, ShanghaiTech, UCSD Ped1, and UCSD Ped2, and achieved AUC scores of 93.4 %, 78.8 %, 93.5 %, and 95.7 % for these datasets, respectively. The experimental results demonstrate the flexibility and generalizability of the proposed method for video anomaly detection, coming from robust features extracted by a pre-trained vision transformer and efficient learning of data representations by employing the AdaBoost training strategy. © 2024 Elsevier Ltd
Applied Soft Computing (15684946) 148
Although the high number of bands in hyperspectral remote sensing images increases their usefulness, it also causes some processing difficulty. In supervised classification, one problem is decreasing classification accuracy due to the insufficient training samples against the bands. A way to deal with this problem is the selection of appropriate bands by the metaheuristic methods. Because of the stochastic search, the selected bands differ in any implementation of a metaheuristic method. So, the results obtained from the classification of these different band subsets will also have some differences. In this study, a fusion-based approach has been proposed to improve the classification of hyperspectral remote sensing images by multiple implementations of a metaheuristic method for band selection. We have tested the proposed method using ten metaheuristic methods with different objective functions on four well-known datasets. The results show the proposed fusion-based approach successfully improves the classification accuracy in all experiments. The accuracy improvement varies depending on the metaheuristic method, the objective function, and the dataset and ranges from 0.4% to 15.7%. The proposed method improves the classification of complex datasets more and affects weaker objective functions considerably. The results also show the proposed method brings the accuracy of different metaheuristic methods close to each other and reduces the sensitivity of selecting the proper ones. Thus, an automated classification system can be obtained using a parameter-less method. © 2023 Elsevier B.V.
IEEE Sensors Journal (1530437X) 23(14)pp. 15570-15577
With strong evanescent waves, optical microfibers (MFs) provide guided lights the ability to directly interact with surrounding environments, whereby several fiber optics chemical sensors have been realized. In this study, based on MFs external refractive index (RI) sensitivity, a multimode optical fiber (MMF) specklegram RI sensor with MMF-MF-MMF configuration is presented. As the MF Section is exposed to the liquids of different RIs (in the range of 1.333-1.368), the interaction between the liquids and evanescent waves modulates the guidance status of the MF, thereby changes the excited modes within the end MMF Section and affects the output speckle pattern accordingly. The evaluation of the functionalities of the MFs with different waist diameters (11, 18, 33, and 42 μ m ) shows that the MF with 33- μ m waist diameter results in the highest output specklegram RI sensitivity, which has been quantified by the zero-mean normalized cross correlation coefficient (ZNCC). Moreover, the response time and sensitivity of the proposed fiber specklegram sensor (FSS) have been simultaneously improved by applying spacial filter on the captured speckles. The RI sensor has also been studied for the temperature detection and showed 0.013°C-1 linear sensitivity within the range of 25 °C-65 °C. Finally, the theoretical analysis of the supported modes by the MFs of the specified waist diameters verifies that 33- μ m sample with high number of guided modes and strong total evanescent waves is the optimum case for the MMF-MF-MMF specklegram RI sensor. © 2001-2012 IEEE.
Infrared small target detection is a challenging task. Despite the recent advances in the development of small infrared target detection algorithms, having a robust target detection algorithm with high probability of detection rate remains unaddressed. To this end, in this paper, a morphological top-hat transform is presented. The proposed method benefits from intrinsic high probability detection rate of classical top-hat transform, while the false alarm rate remains below the acceptable threshold. Also, the sensitivity to noise analysis demonstrates that the proposed method is robust against various noise intensities. The proposed method is tested on some real infrared images. The experiments show that the proposed method outperforms state-of-the-arts in both quantitative and qualitative manners. © 2023
Signal, Image and Video Processing (18631711) 17(7)pp. 3247-3254
In most multi-frame image super-resolution (SR) studies, the process that models the construction of the low-resolution (LR) images from high resolution (HR) one includes geometric transformation, blurring, downsampling, and adding noise. A simple linear approximation of this model is used in most super-resolution frameworks. The approximation error in this popular model appears as an additive noise that is generally a non-Gaussian noise. Most state-of-the-art estimators, such as the Kalman filter as a linear estimator used in super-resolution problems, suffer from this non-Gaussian noise caused by approximation error. We present a novel multi-frame super-resolution using the particle filter method that performs well in a non-Gaussian nonlinear dynamic system to produce the high-resolution output. One key innovation is that the approximation error in the super-resolution model can be handled well by using a particle filter. In this study, the optimal importance function is used to improve particle filter efficiency determined with the Taylor expansion of observation and measurement function. Experiments with simulated and real-image sequences yield that our proposed super-resolution method has good results in suppressing noise and reconstructing details in high-resolution images. © 2022, The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature.
IEEE Sensors Journal (1530437X) 22(13)pp. 13144-13152
Due to the limited resolution of depth maps captured by RGB-D sensors, depth map Super Resolution (SR) techniques have received a lot of attention. Intensity guided depth map SR methods based on bilateral filter or guided image filter are commonly used for depth upsampling. Although promising edge-preserving results have been reported in these methods, texture-copying artifacts caused by structure discrepancy between depth map and associated intensity image cannot be addressed, easily. In this paper we aim to balance the trade-off between preserving structure and suppressing texture defects. Based on this, a structure-preserving guided filter is presented that not only keeps the advantages of aforementioned methods, but also overcomes texture-copying artifacts. Unlike conventional guided filtering-based methods which rely on only one guidance, we emphasize on the use of both intensity and depth information as guidance to alleviate the deficiencies of the existing works. We replace the mean filtering scheme in guided filters with a weighted average strategy, where the weights are described by the local depth kernel depended on the input depth map. This enables our method to considerably reduce texture-copying artifacts while preserving 3D structural details. Visual evaluation of results shows that the algorithm can also avoid halo artifacts near the edges whereas traditional guided filters suffer from it. Quantitative results of comprehensive experiments demonstrate the effectiveness of our approach over prior depth map SR works. © 2001-2012 IEEE.
IEEE Access (21693536) 10pp. 120592-120605
Despite significant advances and innovations in deep network-based vehicle detection methods, finding a balance between detector accuracy and speed remains a significant challenge. This study aims to present an algorithm that can manage the speed and accuracy of the detector in real-time vehicle detection while increasing detector speed with accuracy comparable to high-speed detectors. To this end, the Fast-Yolo-Rec algorithm is proposed. The proposed method includes a new Yolo-based detection network and LSTM-based position prediction networks. The proposed semantic attention mechanism in the spatial semantic attention module (SSAM) significantly impacts accuracy and speed on par with the most recent fast detectors. Recurrent position prediction networks, on the other hand, improve the detection speed by estimating the current vehicle position using vehicle position history. The vehicle trajectories are classified, and the LSTM network for the specified trajectory predicts the vehicle positions. The Fast-Yolo-Rec algorithm not only determines the position of the vehicle faster than high-speed detectors but also allows for the speed control of the detection network with acceptable accuracy. The evaluation results on a large Highway dataset show that the proposed scheme outperforms the baseline methods. © 2013 IEEE.
Geocarto International (17520762) 37(12)pp. 3565-3576
Building recognition is a core task for urban image classification (mapping), especially in optical high-resolution imagery. Convolutional Neural Networks (CNNs) have recently achieved unprecedented performance in the automatic recognition of objects (e.g. buildings, roads, or trees) in high-resolution imagery. Although these results are promising, questions remain about generalizability. This is a great challenge, as there is a wide variability in the visual characteristics of the building image scene across different geographic locations. CNNs are overfitted with limited and low diversity samples and are tested on the same or nearby geographic locations. In this work, we propose two scenarios with regard to transfer learning CNN features for building scene classification. We also investigate the generalizability of CNNs for building recognition across different geographic locations. The results of the two scenarios show that the final model, generalizable in different geographic locations, unseen areas. © 2021 Informa UK Limited, trading as Taylor & Francis Group.
Multimedia Tools and Applications (13807501) 81(8)pp. 11461-11478
Depth images captured by conventional RGB-D sensors such as ToF cameras have limited resolution. Despite recent advances in depth camera technology, there is still a significant difference between the resolution of depth and color images. Therefore, depth map Super-Resolution (SR) techniques have received attention. Specifically, achieving an algorithm performing well at large scaling factors is of great importance and also challenging. In most existing methods, the up-sampling of low resolution depth images to the desired size is performed by an interpolation operation during the beginning stage and quality improvement filters are applied then. Due to the different nature of depth images and their sparsity, magnifying the images in a single step brings heavy artifacts specially at large up-sampling factors (e.g., 16). To tackle this problem, we propose a progressive multi-step depth map SR method where interpolation and modified enhancement processes are applied iteratively. This extremely improves the quality of the output depth image. Moreover, considering the importance of edges and discontinuities in depth images, instead of using conventional symmetric kernel, an edge directed kernel is applied which effectively avoids blurring. In addition, texture copying and depth bleeding artifacts are reduced employing a depth range filter. Quantitative and qualitative results of comprehensive experiments on Middlebury and real-world datasets demonstrate the effectiveness of our approach over prior depth SR works, especially for large scaling factors of 16, 32 and even 64. © 2022, The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature.
This paper presents a novel approach in feature extraction, Local Symmetric Directional Pattern (LSDP), for face recognition. The LSDP encodes the structure of facial textures based on gradient information in a simple and compact coding approach to produce more distinctive code in less time and memory than existing methods. We extract gradient information by convolving the face image with four symmetric compass masks to encode this information using directional numbers, which are related to directional information, and magnitudes of the two prominent edge responses. We also use a hybrid feature vector as a face descriptor obtained by reducing the dimensions of the LSDP feature map and classify them using the sparse representation-based classification (SRC) algorithm. Due to the high discrimination power of the extracted features, the construction of a dictionary based on the hybrid features leads to more sparse representation coefficients. As a result, it improves SRC performance in terms of recognition rate and computational speed. We perform extensive experiments to evaluate and compare the performance of our descriptor with other handcrafted descriptors and the deep learning feature under two different evaluation protocols (different dimensions of feature space and the different number of training samples) on different face databases, which have different variations of illumination, expression, and pose. Experimental results illustrate that our method achieves the highest recognition rate compared to other methods in both evaluation protocols. Especially under challenging conditions where the dimensions of the feature space or the number of training samples are low, LSDP demonstrates excellent performance. © 2021 Elsevier GmbH
This paper proposes a novel descriptor for facial expression recognition called the Local Directional Compact Pattern (LDCP), which encodes the prominent information of local facial textures in a simple and compact way. We obtain edge responses for the local neighborhood of each pixel in four different directions by convolving the symmetric compass masks with the face image. LDCP uses direction, sign, and position information of the two top edge responses to generate a more distinctive code than existing methods that allows us to distinguish different textures with similar gradient directions. Unlike other methods in which the whole face is uniformly divided into several blocks to obtain a global feature vector by concatenating the histogram of each block, we select emotion-related blocks from the feature map to obtain the histogram-based feature vector of each block, which have a different contribution to exhibiting facial expressions. Again, we assign a weight to each block to classify facial expressions using a robust kernel representation algorithm. We conduct our experiments on the CK+, FACES, MMI, and JAFFE datasets to compare our LDCP descriptor performance with 20 existing descriptors in terms of the number of bits, base length, feature extraction time, and recognition rate. In addition, we compare our proposed method with the recent state-of-the-art methods in different testing strategies. © 2022 Elsevier GmbH
A hybrid and robust digital image watermarking is suggested by using advantages of both frequency domain and wavelet domain. In this method, the most suitable color component of the host image and the most efficient wavelet sub-band are chosen, concluding from the comprehensive experiments. In this paper, three parallel stages are used to embed and extract the watermark message. After transforming color component of the host image into the wavelet domain, some preprocessing with Discrete Cosines Transformers are applied, and then the watermark message is embedded into the best range by using a secret key. By detecting the rank of constructed detection matrices and voting among three output bits, the watermark bit is extracted. Simulation results demonstrate that the proposed method, besides the high robustness against Gaussian noise-based attacks especially AWGN, has performed well in imperceptibility and embedding capacity factors. © 2022 Elsevier GmbH
IEEE Geoscience and Remote Sensing Letters (1545598X) 19
Getting the advantages of hyperspectral remote sensing images depends on overcoming the challenges posed by their large number of bands. One approach is selecting appropriate bands by the metaheuristic methods. Initial versions of these methods usually suffer from trapping into local optima, so they do not work optimally in selecting the best band subset. The number of training samples is also important in the band selection. With insufficient training samples, even improved metaheuristic methods do not yield the desired result. However, providing sufficient ground data as training samples is costly. In this letter, an improved Levy flight (LF)-based version of the genetic algorithm (GA) is developed and used to select bands in a semisupervised manner. The number of training samples in the proposed semisupervised method is increased based on spatial adjacency and spectral similarity simultaneously. Our results show that in cases where the initial version of the GA fails to select appropriate bands, our improved LF-based version introduces a band subset that yields the desirable classification result. Also, the classification accuracies have been improved considerably using our proposed semisupervised method. In some of our experiments using a small number of training samples, the accuracy improvement is near 17%. The proposed method has been effective in the case of sufficient training samples too. In this case, the accuracy improvement is near 11% in some experiments. © 2004-2012 IEEE.
Digital Signal Processing: A Review Journal (10954333) 117
Small infrared target localization and tracking are of great importance in early-warning systems. In order to accurately localize the target, a high-performance target detection algorithm is required. In this paper, a new detection algorithm is proposed, which effectively enhances the target area and eliminates noise and background clutter. The algorithm is inspired by the minimum variation directions interpolation. The detection performance of the method is investigated comprehensively in different situations. Also, to exclude the effect of thresholding on the detector's performance, a measure based on constant false alarm rate (CFAR) is employed. Experiments on multiple real-world infrared sequences demonstrate the effectiveness of the proposed method. © 2021 Elsevier Inc.
IET Radar, Sonar and Navigation (17518784) 14(6)pp. 811-821
Sparse representation displays remarkable characteristics when applied to image processing and classification. The critical point in the success of sparse representation-based classification is to learn an authentic dictionary. The present study proposes a virtual dictionary-based sparse representation for automatic target recognition. Based on the properties of the synthetic aperture radar (SAR) images, some low complexity modules including adding speckle noise, histogram equalisation mapping, and bicubic interpolation are applied to construct some virtual compact dictionaries using Fisher discriminative dictionary learning. These dictionaries have different discriminative information on targets, which are used independently in several sparse representation-based classifiers. The reconstruction error vectors of the latter classifiers are then combined to recognise the target using decision fusion. Based on experimental results obtained drawing upon moving and stationary target acquisition and recognition data set, the proposed method presents the highest accuracy in classification reported yet in the literature. Furthermore, the procedure improves the recognition robustness against most commonly extended operating conditions, e.g. speckle noise corruption, depression angle variation and reduced training set. Accordingly, the current study claims a robust parallel method of high real-time ability in the target recognition of SAR images applicable to practical situations. © 2020 The Institution of Engineering and Technology.
Multimedia Tools and Applications (13807501) 79(11-12)pp. 7449-7469
In this paper, we proposed a video error concealment algorithm using Motion Vector (MV) recovery for parallelogram partitions in the lost area. Error concealment is inevitable when some video packets are lost during transmission and correction or retransmission is not feasible. In conventional methods, MVs are recovered for the square shaped blocks which are then used for motion compensated temporal replacement. But in our proposed method, by parallelogram partitioning of the lost area, the MVs are found for more general shaped blocks. The parallelograms with various sizes and angles are examined, and then the best combination (size and angle) is selected with the assist of a border matching algorithm and a blind quality assessment method. Experimental results show that our method outperforms the other error concealment algorithms, both subjectively and objectively. © 2019, Springer Science+Business Media, LLC, part of Springer Nature.
Signal Processing (01651684) 177
Infrared small target detection in an infrared search and track (IRST) system is a challenging task. This situation becomes more complicated when high gray-intensity structural backgrounds appear in the field of view (FoV) of the infrared seeker. While the majority of the infrared small target detection algorithms neglect directional information, in this paper, a directional approach is presented to suppress structural backgrounds and develop a more effective detection algorithm. To this end, a similar concept to the average absolute gray difference (AAGD) is utilized to construct a novel directional small target detection algorithm called absolute directional mean difference (ADMD). Also, an efficient implementation procedure is presented for the proposed algorithm. The proposed algorithm effectively enhances the target area and eliminates background clutter. Simulation results on real infrared images prove the significant effectiveness of the proposed algorithm. © 2020
Electronics Letters (1350911X) 56(12)pp. 619-621
Sparse representation-based classification (SRC) possesses remarkable characteristics for application in synthetic aperture radar (SAR) automatic target recognition (ATR), for instance, inherent feature extraction and robustness to articulation, and so on. However, the performance of SRC is highly sensitive to parameters such as sufficient training samples, SAR images quality, and targets' changing conditions in depression, pose, configuration, and so on. Unfortunately, the training sample resources for SAR ATR are often expensive and scarce. Further, the targets in SAR images, even with slightest changes in conditions, display mutable characteristics attributable to unique SAR image formation, and speckle noise corruption. To overcome these obstacles, this Letter proposes to establish several compact and complementary dictionaries using monogenic signal's components of SAR images and Fisher discriminative dictionary learning. Then, an optimal decision fusion (ODF) strategy is proposed, which utilises SRC and the latter dictionaries for robust SAR ATR. Compared with single classifiers or multiple parallel classifiers, the proposed ODF increases the accuracy of recognition, while at the same time, decreases the complexity of the system. The proposed methods have considerable features making them applicable in practical situations. Based on the experimental results, the proposed methods outperform state-of-the-art approaches. © The Institution of Engineering and Technology 2020
MICROELECTRONICS JOURNAL (00262692) 101
One of the problems with the Controller Area Network is jitter of response time because of the bit-stuffing mechanism. In real-time cases where timing accuracy is important, jitter may deteriorate the quality of the control procedure noticeably. A new method based on XOR masking, named Statistical Mask Calculation (SMC), is presented in this study. The method uses the statistical parameters of data to generate a proper mask for each case. The performance of the proposed method on the reduction of the number of stuff bits is evaluated by considering real data set and a comparison with existing original XOR masking is done. The results indicate that applying the proposed method on the case study increases the probability of zero bit stuffing up to 46% compared to the original XOR mask. It should be noted that the proposed technique is more effective for systems that are usually predictable. Therefore, the adaption of this technique depends on the required application. © 2020 Elsevier Ltd
IEEE Access (21693536) 8pp. 133817-133826
In this paper, a new approach is introduced to reduce bit stuffing and consequently residual error probability in the controller area network (CAN). The proposed method is based on XOR masking. Unlike the XOR method, the proposed approach does not use a fixed mask for all IDs. Using statistical parameters of data, a proper mask for each CAN ID is generated and applied to the messages before transmitting. The performance of the method to reduce bit stuffing and residual error probability has been evaluated by considering a real data set. Results show that the method can significantly reduce bit stuffing and residual error probability. A comparison has been also conducted with previously reported methods. The results show the superiority of the SMC method in reducing residual error without payload and data transfer rate reduction. © 2013 IEEE.
Signal, Image and Video Processing (18631711) 13(5)pp. 895-903
In this paper, we present a new interpolation filter for salt and pepper noise (SPN) restoration in digital images. The proposed filter is established on decision-based filters (DBF) and consists of two units: noise detection and noise restoration. In noise restoration unit, the natural neighbor Galerkin method (NNGM), which is a two-dimensional scattered data interpolation, is adopted for estimating the intensities of noisy pixels. For each detected noisy pixel, an adaptive size local filtering window is considered and NNGM interpolation is applied locally. Compared to the other state-of-the-art DBFs for SPN removal, which are based on interpolation methods, our proposed method has better performance in terms of both objective and subjective assessments. In our proposed method, the noise restoration unit requires no manual parameter tuning. Numerous experiments results demonstrate that our proposed filter removes SPN effectively and preserves the details and edges well. © 2019, Springer-Verlag London Ltd., part of Springer Nature.
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing (21511535) 12(2)pp. 675-684
Although scale invariant feature transform (SIFT) based algorithms have a wide range of applications in remote sensing panchromatic image matching, but minor changes in the contrast threshold can bring about drastic changes in the image matching quality. In order to effectively improve the matching quality of the SIFT detector, this paper proposes an adaptive contrast threshold-SIFT (ACT-SIFT) procedure. The ACT-SIFT method set forth herein harmoniously calculates two sought after contrast thresholds for the target and the reference images separately an objective achieved through minimizing certain proposed criterion. As an introductory step, the entropy in the keypoints (EiK) is defined as a control or balancing parameter being estimated in terms of the entropy of the image and the number of the keypoints. To that end, the candidate keypoints are first extracted from the scale space of SIFT method by applying the initial contrast threshold for both images. Then, the contrast threshold values are modified iteratively for both images to reach the needed values by minimizing the relative difference between the EiK of the target and the reference images. Next, in agreement with the threshold values obtained, the keypoints are extracted with the help of the SIFT keypoint detector. The correct matching pairs are also created using the matches acquired through the SIFT descriptor. The results obtained by applying our proposed approach promise boosted matching pairs in remote sensing image correspondences, for it extracts the keypoints in a robust manner by simultaneously checking the information content of the reference and the target images. © 2018 IEEE.
Signal Processing (01651684) 164pp. 84-98
Although correlation filter (CF)-based trackers have shown promising results in addressing problematic challenges of visual tracking, common holistic-wise CF-based trackers mostly drift away from the target object when undergoing partial occlusion. On the other hand, part-based models provide a prosperous basis for handling occlusion problem, due to preserving local structure of the target object. Employing local-global appearance models of the object, we propose a robust tracking algorithm based on the weighted cumulative fusion of CF-based part regressors. Indeed, we dynamically learn importance weights of each part via a multilinear ridge regression optimization model aiming at enhancing discrimination power of our tracker. To alleviate tracking drift caused by the object size changes, we further present an accurate method that jointly estimates object scale and aspect ratio by analyzing relative deformation cost of importance pair-wise parts. Also, to reduce the computational complexity, we introduce a feature sharing strategy for all constituent parts. Extensive experiments on OTB-2013, OTB-50, OTB-100, and VOT2016 datasets demonstrate that our tracker not only impressively enhances the performance of target-wise KCF tracker as its baseline but also performs favorably against state-of-the-art trackers in terms of qualitative and quantitative measures while running about 30 fps using Matlab on 3.2 GHz core-i5. © 2019 Elsevier B.V.
Asiri, S. ,
Khademianzadeh, F. ,
Monadjemi, A. ,
Moallem, P. IEEE/ASME Transactions on Mechatronics (10834435) 24(5)pp. 2406-2415
This article discusses the design, development, and energy consumption of a power-efficient spherical robot. The boosted performance of the robot is due to the separation between the lateral and longitudinal motions. This two-pendulum Innovative Spherical Robot (ISR) is 11% more energy efficient than those of similar one-pendulum models. The complex and nonlinear dynamic model of this nonholonomic robot is accurately derived by employing the Euler-Lagrange method from three considered points of masses. By resorting to experimental tests, the results of the dynamic models simulations are presented and verified. Through some carefully designed and carried-out experiments, the maneuverability of the ISR is demonstrated. Some specific curvature lanes are devised to measure the energy consumption of the model. As indicated by the experimental results, the maximum cruise time obtained by the ISR is 220 min, which means 7 km range that suggests an appreciable ability for the long-term runs. © 2019 IEEE.
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing (21511535) 12(8)pp. 2627-2635
Automated building extraction from single, very-high-resolution (VHR) images is still one of the most challenging tasks for urban planning, estimating population, understanding urban dynamics, and many other applications. The complexities of building objects have caused the images of buildings to be oversegmented into multiple segments in the object-based image analysis (OBIA) method. Selecting the appropriate segmentation scale parameter is a major challenge in OBIA that influences the discriminative features extraction, especially for building objects. Furthermore, transferability of OBIA method is another challenge. Presently, convolutional neural networks (CNNs) are a well-understood tool for images scene classification. However, scene classification based on CNNs is still difficult due to the scale variation of the objects in VHR images. To meet these challenges, we propose a novel object-based deep CNN (OCNN) framework for VHR images. The datasets used for testing were Vaihingen (Germany) aerial images and a Tunis Worldview-2 (WV2) satellite imagery. Experimental results prove that our framework is extensible to different types of the image with the same sensor or another sensor (for example WV2) with once-fine-tuning. In addition, our framework extracts the different types of building with respect to size, color, material, spectral similarity to roads, and complex backgrounds. Quantitative evaluation at the object level demonstrated that the proposed framework could yield promising results (average precision 0.88, recall 0.92, quality 0.82, F-score 0.90, overall accuracy 0.95, and Kappa coefficient 0.90). Comparative experimental results indicate that our proposed OCNN significantly outperforms the traditional method for building extraction. © 2008-2012 IEEE.
Signal Processing (01651684) 155pp. 108-129
Kalman filter (KF) as a linear estimator which is used in super-resolution (SR) problems, suffers from high computational costs and storage requirements. To gain appreciable success in the elimination of these two challenges, this paper advances a SR framework employing KF in the frequency domain, while no resort is made to any approximations or extra assumptions in the dynamic system modeling and statistical matrices. Generally, previous KF-based SR methods organized the system with huge-sized matrices in the spatial domain, following which they tried to reduce the system dimension using approximation and/or limitation on point spread function (PSF). In this study, first, several small-dimension dynamic systems are separately made in the frequency domain supporting space-invariant PSFs of an arbitrary form and size. Then, the acquired small-dimension KF estimators are applied rather than the traditional huge-dimension one. These will greatly reduce computational complexity, decrease storage requirements allowing parallel implementation as well. Furthermore, our proposed SR framework can be used to produce high resolution image of an expedient size, that is, a scalable SR. Experimental results with both simulated and real world sequences indicate that our proposed framework works more effectively than the other compared methods, especially in fine details restoration. © 2018 Elsevier B.V.
Infrared Physics and Technology (13504495) 89pp. 387-397
False alarm rate and detection rate are still two contradictory metrics for infrared small target detection in an infrared search and track system (IRST), despite the development of new detection algorithms. In certain circumstances, not detecting true targets is more tolerable than detecting false items as true targets. Hence, considering background clutter and detector noise as the sources of the false alarm in an IRST system, in this paper, a false alarm aware methodology is presented to reduce false alarm rate while the detection rate remains undegraded. To this end, advantages and disadvantages of each detection algorithm are investigated and the sources of the false alarms are determined. Two target detection algorithms having independent false alarm sources are chosen in a way that the disadvantages of the one algorithm can be compensated by the advantages of the other one. In this work, multi-scale average absolute gray difference (AAGD) and Laplacian of point spread function (LoPSF) are utilized as the cornerstones of the desired algorithm of the proposed methodology. After presenting a conceptual model for the desired algorithm, it is implemented through the most straightforward mechanism. The desired algorithm effectively suppresses background clutter and eliminates detector noise. Also, since the input images are processed through just four different scales, the desired algorithm has good capability for real-time implementation. Simulation results in term of signal to clutter ratio and background suppression factor on real and simulated images prove the effectiveness and the performance of the proposed methodology. Since the desired algorithm was developed based on independent false alarm sources, our proposed methodology is expandable to any pair of detection algorithms which have different false alarm sources. © 2018 Elsevier B.V.
Journal of Circuits, Systems and Computers (17936454) 27(2)
Smart image sensors with low data rate output are well fitted for security and surveillance tasks, since at lower data rates, power consumption is reduced and the image sensor can be operated with limited energy resources such as solar panels. In this paper, a new data transfer scheme is presented to reduce the data rate of the pixels which have undergone value change. Although different pixel difference detecting architectures have been previously reported but it is shown that the given method is more effective in terms of power dissipation and data transfer rate reduction. The proposed architecture is evaluated as a 100×160-pixel sensor in a standard CMOS technology and comparison with other data transfer approaches is performed in the same process and configuration. © 2018 World Scientific Publishing Company.
Journal of Visual Communication and Image Representation (10473203) 51pp. 40-55
Intensity restoration of pixels corrupted by impulse-noise contributes greatly to the quality of decision based filters (DBF). In this paper, we present an efficient structural post-processing method, which is based on directional-correlation, linear-regression-analysis, and inverse-distance-weighted-mean techniques. The proposed method is adopted as a complementary part after DBFs to enhance the quality of the final restored image. We assume that by adopting the preliminary DBF, noisy-pixels are detected by noise-detection unit and afterwards their intensities are estimated by the noise-restoration unit. In our method for each detected noisy-pixel, the intensity variation of adjacent pixels of restored image on different directions are analyzed in the corresponding local window and based on this structural information, the intensity of the previously-restored noisy-pixel is modified more accurately. Since the structures in images are more recognizable for low-density impulse-noise, our method is more effective in this case however a gradual improvement is achieved for high-density impulse-noise. © 2017 Elsevier Inc.
Digital Signal Processing: A Review Journal (10954333) 75pp. 242-254
In recent years, decision based filters (DBFs) are the most popular technique for impulse-noise restoration. The DBFs consist of two stages: noise-detection and noise-restoration. The performance of noise-restoration stage affects the quality of DBFs significantly. In this paper, we presented an effective structural based refinement method which could be adopted as a complementary stage after DBFs to improve the quality of the final restored image. Here, we assume that the preliminary DBF has detected the noisy-pixels and has restored the intensities of the noisy-pixels. In our proposed refinement method for each detected noisy-pixel, based on local structural information of the image, the previously restored intensity of noisy-pixel is modified more accurately. This is performed by analyzing the gradient of output restored image of preliminary DBF and calculating direction of contour which are passed through the noisy-pixels. Then based on the angular difference of contour-direction with 4 main lines, which are passing through the noisy-pixel, the previously restored intensity of noisy-pixel is replaced with weighted means of surrounding pixels' intensities. Since the structures in images are more recognizable for low-density impulse-noise, our method is more effective in this case, however a small improvement is obtained for high-density impulse-noise. © 2018 Elsevier Inc.
International Journal on Engineering Applications (25332295) 6(1)pp. 29-34
Automatic diagnosis of diseases always has been of interest as an interdisciplinary study amongst computer and medical science researchers. In this paper, application of artificial neural networks in typical disease diagnosis has been investigated. The real procedure of medical diagnosis which usually is employed by physicians was analyzed and converted to a machine implementable format. Then after selecting some symptoms of eight different diseases, a data set contains the information of a few hundred cases was configured and applied to a MLP neural network. The results of the experiments and also the advantages of using a fuzzy approach were discussed as well. Outcomes suggest the role of effective symptoms selection and the advantages of data fuzzification on a neural networks-based automatic medical diagnosis system. Employing ETM diseases as the case study, system eventually gets through the 97.5% of correct detection of abnormal cases. © 2018 Praise Worthy Prize S.r.l.-All rights reserved.
Digital Signal Processing: A Review Journal (10954333) 72pp. 19-43
As there are data redundancies in successive frames in a multi-frame super resolution (SR) algorithm, one can expect that discarding some of these superfluous frames would have no impact on the quality of the high resolution (HR) output image. The present paper presents an efficient algorithm for selecting the proper combination of the minimum frames required for multi-frame SR algorithms so as to not only preserve the quality of the obtained HR output, but also reduce the SR procedure complexity and memory. To achieve this, the present study first seeks to prove that minimizing the spectral interference between the selected frames for SR procedure will result in maximizing the HR output power. Then, the criterion for measuring the Upper Bound on Spectral Interferences (UBSI) among the selected frames for SR procedure is presented; the formulation is expressed in such a way that it can be extended to global sub-pixel translations between frames. Our proposed frame selection algorithm evaluates all candidate combinations from input frames so that the best option capable of minimizing the UBSI can be selected. In order to evaluate our proposed frame selection algorithm, five well-known SR image reconstruction methods are applied both in four standard simulated images and in three well known real video sequences, employing two different procedures: Using our proposed frame selection algorithm and otherwise. The obtained results indicate that when our proposed frame selection algorithm is applied, the quality of the HR output images is preserved tantamount to considering all available frames. Besides, the computational complexity of the SR algorithms is dramatically reduced adopting the proposed frame selection algorithm, for the number of frames engaged in the SR is diminished. Also compared with the SR algorithms presented in the literature, our proposed frame selection method takes relatively negligible time to execute. © 2017 Elsevier Inc.
IET Image Processing (17519667) 12(9)pp. 1577-1585
Restoring pixel intensities corrupted by impulse noise has a great impact on the quality of decision-based filters. In this study, the authors' focus is on intensity restoration of noisy pixels. Their assumption is that noisy pixels are already established by the noise-detection unit being considered as missing data in the image. When the interpolation methods are adopted in the noise-restoration unit of the decision-based filters for the purpose of restoring the intensities of the noisy pixels, two unexpected problems emerge - jagged edges and blurred details. These drawbacks can be ameliorated by using extra information obtained from structures in the images. Their structure-based interpolation method comprises two steps: preinterpolation and post-interpolation. In the first step (pre-interpolation), the Sibson natural neighbour interpolation is adopted for the initial estimation of the intensities of all noisy pixels. In the second step (post-interpolation, modifying-phase), for each noisy pixel in pre-interpolated image, the intensity variations of the pixels on two adjacent parallel lines, in different directions in their corresponding local windows, are analysed. Based on the obtained structural information, the intensity of the centred noisy pixel is restored more effectively. Since the structures in the images are far more noticeable at low-density impulse noise, the proposed method works more efficiently in this case; however, a gradual improvement is achieved for high-density impulse noise. © 2018, The Institution of Engineering and Technology.
Signal, Image and Video Processing (18631711) 11(6)pp. 1009-1016
Intravascular ultrasound (IVUS) is clinically available for visualizing coronary arteries. However, it suffers from acoustic shadow areas and ring-down artifacts as two of the common issues in IVUS images. This paper introduces an approach which can overcome these limitations. As shadow areas were displayed behind hard plaques in the IVUS grayscale images, calcified plaques were first segmented by using Otsu threshold. Then, active contour, histogram matching, and local histogram matching are implemented. In addition, a new modified circle Hough transform is introduced to remove the ring-down artifacts from IVUS images. In order to evaluate the efficacy of this new method in detection of shadow and ring-down regions, 300 IVUS images are considered. Sensitivity of 89% and specificity of 92% are achieved from a comparison in revelation of calcium along with shadow in the proposed method and virtual histology images. Also, area differences of 5.83 ± 3.3 and 5.65 ± 2.83 are obtained, respectively, for ring-down and shadow domain when compared to measures performed manually by a clinical expert. © 2017, Springer-Verlag London.
Information Processing in Agriculture (20970153) 4(1)pp. 33-40
In this paper, a computer vision-based algorithm for golden delicious apple grading is proposed which works in six steps. Non-apple pixels as background are firstly removed from input images. Then, stem end is detected by combination of morphological methods and Mahalanobis distant classifier. Calyx region is also detected by applying K-means clustering on the Cb component in YCbCr color space. After that, defects segmentation is achieved using Multi-Layer Perceptron (MLP) neural network. In the next step, stem end and calyx regions are removed from defected regions to refine and improve apple grading process. Then, statistical, textural and geometric features from refined defected regions are extracted. Finally, for apple grading, a comparison between performance of Support Vector Machine (SVM), MLP and K-Nearest Neighbor (KNN) classifiers is done. Classification is done in two manners which in the first one, an input apple is classified into two categories of healthy and defected. In the second manner, the input apple is classified into three categories of first rank, second rank and rejected ones. In both grading steps, SVM classifier works as the best one with recognition rate of 92.5% and 89.2% for two categories (healthy and defected) and three quality categories (first rank, second rank and rejected ones), among 120 different golden delicious apple images, respectively, considering K-folding with K = 5. Moreover, the accuracy of the proposed segmentation algorithms including stem end detection and calyx detection are evaluated for two different apple image databases. © 2017 China Agricultural University
Iranian Conference on Machine Vision and Image Processing, MVIP (21666776) 2017pp. 114-118
In fiber optic sensors, laser source produces a coherent light which is transmitted through a fiber cable. The output of this laser forms a speckle. This speckle is captured by a CCD and then analyzed by a PC. The intensity and shape of the speckle change when external force is applied on the cable. In this paper, Laplacian of Gaussian (LoG) filter is used to improve the dynamic range and accuracy of the fiber optic sensor. Capturing images is done in a dark room and the simulation is done by Matlab. The difference between the intensities of a normal and abnormal state image was 4.0556 when using 2D Discrete Wavelet transform (DWT) filter. It is increased into 14.6191 when LoG filter is used, thus increasing the dynamic range and accuracy of the sensor. In addition, the time is decreased by about 24.5 %. The number of changed blocks is increased by about 22 % leading to double increase in the accuracy and dynamic range. © 2017 IEEE.
IEEE Transactions on Power Electronics (08858993) 32(4)pp. 2964-2975
In this paper, a new advanced deadbeat direct torque and flux control (A-DB-DTFC) system is proposed that improves the steady-state and transient-state performances of the permanent-magnet synchronous motor by adopting two improved deadbeat methods. Whenever the error between the torque and its reference value is low, an improved deadbeat method is adopted by the A-DB-DTFC system, in which the phase and time duration of the voltage vector applied to the motor are adjusted in a manner that the stator flux and torque reach their reference values after just one control cycle. Whenever the torque error is high, another deadbeat method is adopted by the A-DB-DTFC system, where the voltage vector phase is adjusted such that the fastest torque response is achieved. In order to assess the effectiveness of the proposed A-DB-DTFC system, the steady-state and transient-state performances of the motor are tested in MATLAB software and in practice, where the simulation and experimental results confirm that the proposed control system reduces the torque and stator flux ripples and achieves the fastest dynamic response. The comparative assessment with the recent DB-DTFC method indicates that the proposed A-DB-DTFC system yields lower torque and flux ripples and a faster dynamic response with the advantage of a lower computation complexity. © 1986-2012 IEEE.
International Journal of Tomography and Simulation (discontinued) (23193336) 30(1)pp. 105-117
Active contour models (ACM) with level set evolution equations can be used successfully in segmentation of textured objects in a textured background, which is a challenge in image segmentation field. In this paper, a geometric ACM using color and Gabor features is proposed for textured image segmentation. The different orientations and scales of Gabor features are tested to select the major Gabor coefficients as texture features in our proposed geometric ACM. Moreover, a clustering method is also used to select the best color components of different conventional color spaces. Comparing to other ACM-based texture segmentation methods, our proposed ACM improves the accuracy of segmentation results in texture color images. © 2017 by Ceser Publications.
International Journal of Artificial Intelligence (09740635) 15(1)pp. 163-179
Echo State Networks (ESN) are a special form of recurrent neural networks (RNNs), which allow for the black box modeling of nonlinear dynamical systems. A unique feature of an ESN is that a large number of neurons (the “reservoir”), whose synaptic connections are generated randomly, is used in such that only the connections from the reservoir to the output modified by learning. The computation of optimal weights can then be achieved by a simple linear regression in an offline manner. ESNs have been applied to a variety of tasks from time series prediction to dynamic pattern recognition with great success. In many tasks, however, an online adaptive learning of the output weights is required. Harmony Search (HS) algorithm shows good performance when the search space is large. Here we propose HS algorithm for training echo state network in an online manner. In our simulation experiments, the ESNs are trained for predicting of three different time series including Mackey-Glass, Lorenz chaotic and Rossler chaotic time series with four different algorithms including Recursive Least Squares (RLS-ESN), Particle Swarm Optimization (PSO-ESN), and our proposed methods (HS-ESN and HS-RLS-ESN). Simulation results show that HS-ESN is significantly the fastest algorithm for training ESN whereas can effectively meet the requirements of the output precision. HS-RLS-ESN algorithm firstly uses HS to close to solution region then it uses RLS to obtain less error. HS-RLS-ESN is slower than HS-ESN and faster than RLS-ESN, but its generality power is very close to RLS-ESN. © 2017 [International Journal of Artificial Intelligence].
IEEE Transactions on Power Electronics (08858993) 31(5)pp. 3738-3753
In this paper, a new predictive direct torque control (DTC) method is proposed, improving the dynamic response of the classical DTC and reducing the torque- and flux-ripples through a voltage vector with an optimal phase. In the transient state, the voltage vector phase is selected in a manner where the fastest dynamic response is achieved, while in the steady state, this phase is selected in a manner where the stator flux amplitude reaches its commanding value at the end of the control cycle. In the steady state, the selected vector is applied to the motor with an optimal time duration calculated to achieve the minimum torque ripple. The five-segment space-vector modulation is used to synthesize the voltage vector, where a fixed switching frequency is obtained. To investigate the effectiveness of the proposed method, steady-state and transient-state performances are tested in the MATLAB software and in practice. Both the simulation and experimental results confirm that the proposed method reduces the torque and flux ripples effectively while improving the dynamic response of the classical DTC method. The comparative investigation of the proposed method with the recent DTC methods indicates that the proposed method has lower ripples in the steady state and a faster dynamic response in the transient state. © 2015 IEEE.
International Journal of Remote Sensing (13665901) 37(21)pp. 5234-5248
Building extraction from high-resolution satellite images (HRSI) in urban areas is an intricate problem. Recent studies proposed different methods during 2005–2015. However, in HRSI, they have not investigated the effects of challenges altogether. This paper studies the effects of non-building features which are the main drawbacks in building extraction. To overcome each challenge, it reviews recent strategies between 2005 and 2015. The pros and cons of each strategy are discussed, and proper strategies are combined to generate hybrid methods. Lower cost and fewer strategies are efficient attributes to recognize the best hybrid methods. Hybrid methods can be useful for different case studies in the future. © 2016 Informa UK Limited, trading as Taylor & Francis Group.
IEEE Transactions on Industrial Electronics (02780046) 63(6)pp. 3876-3888
In this paper, a new predictive direct torque control (DTC) method is proposed, which improves the dynamic response of the conventional DTC in the transient state and yields the minimum torque and flux ripples in the steady state through an optimal voltage vector. At the steady state, the magnitude, phase, and time duration of the voltage vector are adjusted in a manner where the minimum torque and flux ripples are obtained; whereas in the transient state, the voltage vector parameters are adjusted in a manner where the fastest dynamic response is achieved. The space-vector modulation is used in synthesizing the selected voltage vector where a fixed switching frequency is achieved. In order to improve the control system efficiency, the principle of maximum torque per ampere is adopted in obtaining the commanding stator flux magnitude. To investigate the effectiveness of the proposed method, the steady-state and transient-state performances are tested in MATLAB software and in practice. The simulation and experimental results confirm that the proposed method yields the minimum torque and flux ripples while improving the dynamic response of the conventional DTC. The comparative investigation with an existing predictive DTC method indicates that the proposed method has a better performance in both the steady state and transient state. © 2015 IEEE.
International Journal of Tomography and Simulation (discontinued) (23193336) 29(3)pp. 32-47
This paper presents an efficient for emotion recognition under occlusion mode of frontal facial images. The proposed algorithm firstly uses combination of Viola-Jones algorithm with skin color information for pure face detection. Then, in the detected face region, the proposed algorithm extracts an optimized Pyramid Histogram of Oriented Gradient (PHOG) descriptor that includes 4 pyramid levels and 4 bins for histogram. The proposed algorithm finally uses a KNN (K Nearest Neighbor) multi-classifier with Euclidean distance which results in high recognitions rate over large databases. The experiments over the RaFD face database show that the average recognition rate of the proposed algorithm for detecting seven common emotions (happy, sad, disgust, fear, angry, surprise and neutral emotions) is 99.52% for non-occluded condition. Moreover, the average recognition rates of the proposed algorithm on JAFFE and CK+, which are another popular face databases, are more than 94.6% and 99.1%, respectively. Averagely, the proposed algorithm achieves to 97.73% emotion recognition rate on over about 2000 facial images. On the other hand, when only half of face image is considered as the system input, the average recognition rate achieves to more than 88.57%, 85.5%, and 96.7% over RaFD, JAFEE and CK+ face images, respectively. Since the proposed algorithm shows high robustness against 50% occlusion in input face images, this algorithm can be used in occlusion conditions too. © 2016 by CESER PUBLICATIONS.
Signal, Image and Video Processing (18631711) 10(2)pp. 351-358
Active contour models (ACM) as deformable shape models are one of the popular methods in object detection and image segmentation. This article presents a robust texture-based segmentation method using parametric ACM. In the proposed method, the energy function of the parametric ACM is modified by adding texture-based balloon energy, so the accurate detection and segmentation of textured object in textured background would be achieved. In this study, texture features of contour, object, and background points are calculated by Gabor filter bank. Then, comparing the calculated texture features of contour points and target object obtains movement direction of the balloon, whereupon active contour curves are shrunk or expanded to make the contour fit to object boundaries. The comparison between our proposed segmentation method and the ACM based on the directional Walsh– Hadamard features, fast adaptive color snake model, and parametric texture model based on joint statistics of complex Wavelet coefficients, indicates that our method is more effective, accurate, and faster for texture image segmentation especially when the textures are irregular or texture direction of object and background is similar. © 2015, Springer-Verlag London.
Infrared Physics and Technology (13504495) 77pp. 27-34
Small target detection is one of the major concern in the development of infrared surveillance systems. Detection algorithms based on Gaussian target modeling have attracted most attention from researchers in this field. However, the lack of accurate target modeling limits the performance of this type of infrared small target detection algorithms. In this paper, signal to clutter ratio (SCR) improvement mechanism based on the matched filter is described in detail and effect of Point Spread Function (PSF) on the intensity and spatial distribution of the target pixels is clarified comprehensively. In the following, a new parametric model for small infrared targets is developed based on the PSF of imaging system which can be considered as a matched filter. Based on this model, a new framework to boost model-based infrared target detection algorithms is presented. In order to show the performance of this new framework, the proposed model is adopted in Laplacian scale-space algorithms which is a well-known algorithm in the small infrared target detection field. Simulation results show that the proposed framework has better detection performance in comparison with the Gaussian one and improves the overall performance of IRST system. By analyzing the performance of the proposed algorithm based on this new framework in a quantitative manner, this new framework shows at least 20% improvement in the output SCR values in comparison with Laplacian of Gaussian (LoG) algorithm. © 2016 Elsevier B.V.
CTIT workshop proceedings series (16821750) 2016pp. 941-946
Anyone knows that sudden catastrophes can instantly do great damage. Fast and accurate acquisition of catastrophe information is an essential task for minimize life and property damage. Compared with other ways of catastrophe data acquisition, UAV based platforms can optimize time, cost and accuracy of the data acquisition, as a result UAVs' data has become the first choice in such condition. In this paper, a novel and fast strategy is proposed for registering and mosaicking of UAVs' image data. Firstly, imprecise image positions are used to find adjoining frames. Then matching process is done by a novel matching method. With keeping Sift in mind, this fast matching method is introduced, which uses images exposure time geometry, SIFT point detector and rBRIEF descriptor vector in order to match points efficiency, and by efficiency we mean not only time efficiency but also elimination of mismatch points. This method uses each image sequence imprecise attitude in order to use Epipolar geometry to both restricting search space of matching and eliminating mismatch points. In consideration of reaching to images imprecise attitude and positions we calibrated the UAV's sensors. After matching process, RANSAC is used to eliminate mismatched tie points. In order to obtain final mosaic, image histograms are equalized and a weighted average method is used to image composition in overlapping areas. The total RMSE over all matching points is 1.72 m.
Signal, Image and Video Processing (18631711) 9(5)pp. 1179-1191
Digital images can suffer from periodic noise, resulting in the appearance of repetitive patterns on the image data and quality degradation. In order to effectively reduce the periodic noise effects, a novel adaptive Gaussian notch filter is proposed in this paper. In the presented method, the frequency regions that correspond to noise are determined by applying a segmentation algorithm on the spectral band of the noisy image using an adaptive threshold. Then, a region growing algorithm tries to determine the bandwidth of each periodic noise component separately. Subsequently, proper Gaussian notch filters are used to decrease the periodic noises only at the contaminated noise frequencies. The proposed filter and some other well-known filters including the frequency domain mean and median filters and also the traditional Gaussian notch filter are compared to evaluate the effectiveness of the approach. The results in different conditions show that the proposed filter gains higher performance both visually and quantitatively with lower computational cost. Furthermore, compared with the other methods, the proposed filter does not need any tuning and parameter adjustments. © 2013, Springer-Verlag London.
Analog Integrated Circuits and Signal Processing (09251030) 85(3)pp. 505-512
In this paper, two various applications of elliptic discrete Fourier transform type I (EDFT_I) are presented in the communication area. In the first application, EDFT_I is applied to reduce the additive uniform and Gaussian noise in the sinusoidal signal. The noise reduction is independent from the type of noise and the corresponding amplitude. In the second application, an EDFT_I-based receiver has been proposed which improves the signal to noise at least about 2 dB for the same error-probability as compared with the optimum receiver considering an additive non-Gaussian noise. In this approach, a binary orthogonal signaling is created using the EDFT_I, which cosine and sine signals are used as carrier for 0 and 1 information. Moreover, for an additive non-Gaussian noise, the proper choice of this transform’s parameters as well as the decision threshold, results in improving the accuracy of digital information’s transmission. © 2015, Springer Science+Business Media New York.
International Journal on Smart Sensing and Intelligent Systems (11785608) 8(3)pp. 1443-1463
Capacitive differential pressure sensor (CPS), which converts an input differential pressure to an output current, is extremely used in different industries. Since the accuracy of CPS is limited due to ambient temperature variations and nonlinear dependency of input and output, compensation is necessary in industries that are sensitive to pressure measurement. This paper proposes a framework for designing of CPS compensation system based on Multi Layer Perceptron (MLP) neural network. Firstly, a test bench for a sample popular CPS is designed and implemented for data acquisition in a real environment. Then, the gathered data are used to train different MLPs as CPS compensation system which inputs are the output current of CPS and temperature value, and the output is compensated current or computed pressure. The experimental results for an ATP3100 smart capacitive pressure transmitter show the trained three layers MLP with Levenberg-Marquardt learning algorithm could effectively compensate the output against variation of temperature as well as nonlinear effects, and reduce the pressure measurement error to about 0.1% FS (Full Scale) , over the temperature range of 5 ~ 60 ° C.
International Journal of Artificial Intelligence (09740635) 13(2)pp. 73-88
A human face detection method for color images is presented in this paper, which is pose, size and position independent, and has the priority of classifying detected faces in three groups: frontal, near frontal and profile, according to their pose. This system is a fuzzy rule base one, optimized by genetic algorithm. In the first stage, skin color regions are selected in the input image. Within each skin area, lip pixels and ear texture are searched, and applied as features to identify face candidates in the skin regions. Summarizing all obtained information along by skin region shape and lip area position relative to skin area, four inputs are computed for fuzzy inference system, and face areas as well as their poses would be introduced. The proposed method is tried on various databases, including HHI, Champion, Caltech, Bao and IMM databases. Achieved results show a remarkable detection rates compared to other methods, for various face poses. 96.8%, 95.3% and 87.8% correct detection rates are achieved, respectively for frontal, near frontal and profile face images over 1298 face image samples. © 2015 by IJAI.
Signal Processing: Image Communication (09235965) 36pp. 95-105
Abstract Multiple Description Coding (MDC) is a technique where multiple streams from a video source are generated, each individually decodable and mutually refinable. MDC is a promising solution to overcome packet loss in video transmission over noisy channels, particularly for real-time applications in which retransmission of lost information is not practical. The error resiliency feature of MDC is achieved at the cost of redundancy, and the required amount of redundancy for each frame depends on the packet loss ratio and also the importance of the frame in the sequence. Due to the error propagation in video transmission over lossy channels, reference frames of a Group of Pictures (GOP) are more important for video reconstruction and, hence, need more redundancy to increase the chance of being received correctly. Therefore a channel adaptive optimization for frame-wise redundancy allocation is needed. In this paper, based on the difference of the side and central decoder outputs, the receiver side distortion is formulated and then used for optimization of a MDC scheme. The performance of the optimizer is verified by experimental results measured from JM 16.0, H.264/AVC reference software. © 2015 Elsevier B.V.
International Review of Aerospace Engineering (19737459) 7(4)pp. 134-141
In this paper a comprehensive model of a tracker system including derive subsystem is presented. Since the main part of the under consideration system consists a two Degree-Of-Freedom (DOF) gimbal subsystem, at first, the model of a two-axis gimbal is considered. For this purpose, after introducing the coordinate systems and different transformation matrices, the related equations of pitch and yaw axes by considering the friction torques, mass unbalances and cable restraint torques are derived. Moreover, all disturbance terms and the methods for their reduction or elimination are investigated. Then, the derive system of the gimbal including a DC motor with a gear is modeled and governing overall equations of the whole system is obtained. In order to have a model with practical considerations, the model of rate gyro which is used for measuring the angular velocities is also included. Finally, the simulation results of the under consideration case study by using the derived comprehensive dynamical equations of this paper are provided. Analysis of these results are used to synthesis the behavior of the practical system and choosing the suitable control structure in order to achieve stabilization and tracking objectives. © 2014 Praise Worthy Prize S.r.l. All rights reserved.
Moallem, P. ,
Amini, F. ,
Habibi M. ,
Amini, F. ,
Habibi, M. ,
Moallem, P. 2025 29th International Computer Conference, Computer Society of Iran, CSICC 2025 pp. 256-260
Retinal image is one of the robust and accurate biometrics which can be used to authenticate an individual. Feature matching is a key step for any biometric system and its implementation on hardware structures is often challenging due to the required object based processing. This paper presents an approach for retina tree biometric matching which has the capability to be implemented on a low power and high speed VLSI hardware. The key idea behind the presented method is to extract the Gaussian profile of the retinal feature dataset. The proposed technique is evaluated on the public VARIA retina image database. © 2014 IEEE.
International Journal of Tomography and Simulation (discontinued) (23193336) 25(1)pp. 101-115
In this paper, a robust stereo matching technique using growing component seeds has been proposed in order to find a semi-dense disparity map. In the proposed algorithm only a fraction of the disparity space is checked to find a semi-dense disparity map. The method is a combination of both the feature based and the area based matching algorithms. The features are the non-horizontal edges extracted by the proposed feature detection algorithm. In order to compute the disparity values in non-feature points, a feature growing algorithm has been applied. The quality of correspondence seeds influences the computing time and the quality of the final disparity map. We show that the proposed algorithm achieves similar results as the ground truth disparity map. The proposed algorithm has been further compared with box filtering, belief propagation, random Growing-Component-Seeds and Canny-Growing-Component-Seeds. According to the obtained results, the proposed method increases the speed of computation ratio of the random GCS algorithm by 31% and the ratio of canny GCS algorithm by 13%. The experimental results show the proposed method achieves higher speed, more accurate disparity map, and the lowest RMS errors. © 2014 by CESER PUBLICATIONS.
International Journal of Power and Energy Systems (10783466) 34(3)pp. 91-98
Power quality monitoring of many sites of an electric power system produces an enormous amount of unstructured data covering the various types of power quality indices. The collected data are not in a suitable form to give insights to the general power quality conditions of a particular site or a particular area within the network. This paper proposes a global power quality index (PQI) which is based on data mining and pattern classification approaches. Firstly, the continuous and discrete PQIs are annually analyzed, normalized and merged. Then, the analyzed indices are classified according to their cost coefficient, and the power quality levels for all distribution sites are determined using the Fast-independent component analysis (ICA) data mining algorithm. Finally, an application example for future explanations is presented.
International Journal of Remote Sensing (13665901) 35(13)pp. 5094-5119
In studies of high-resolution satellite (HRS) imagery, the extraction of man-made features such as roads and buildings has become quite attractive to the photogrammetric and remote-sensing communities. The extraction of 2D images from buildings in a dense urban area is an intricate problem, due to the variety of shapes, sizes, colours, and textures. To overcome the problem, many case studies have been conducted; however, they have frequently contained isolated buildings with low variations of shapes and colours and/or high contrast with respect to adjacent features. As an alternative, this study uses continuous building blocks along with high variation in shape, colour, radiance, size, and height. In addition, some non-building features include either the same or similar materials to that of building rooftops. Thus, there is low contrast between building and non-building features. The core components of the algorithm are: (1) multispectral binary filtering, (2) sub-clustering and single binary filtering, (3) multi-conditional region growing, and (4) post-processing. This approach was applied to a dense urban area in Tehran, Iran, and a semi-urban area in Hongshan district, Wuhan city, central China. A quantitative comparison was carried out between the proposed and three other algorithms for the Wuhan case study. GeoEye multispectral imagery was used in both case studies. The results show that the proposed algorithm correctly extracted the majority of building and non-building features in both case studies. The short running time of this algorithm along with precise manual editing can generate accurate building maps for practical applications. © 2014 Taylor & Francis.
Journal Of Medical Signals And Sensors (22287477) 4(4)pp. 281-290
In order to distinguish between benign and malignant types of pigmented skin lesions, computerized procedures have been developed for images taken by different equipment that the most available one of them is conventional digital cameras. In this research, a new procedure to detect malignant melanoma from benign pigmented lesions using macroscopic images is presented. The images are taken by conventional digital cameras with spatial resolution higher than one megapixel and by considering no constraints and special conditions during imaging. In the proposed procedure, new methods to weaken the effect of nonuniform illumination, correction of the effect of thick hairs and large glows on the lesion and also, a new threshold-based segmentation algorithm are presented. 187 features representing asymmetry, border irregularity, color variation, diameter and texture are extracted from the lesion area and after reducing the number of features using principal component analysis (PCA), lesions are determined as malignant or benign using support vector machine classifier. According to the dermatologist diagnosis, the proposed processing methods have the ability to detect lesions area with high accuracy. The evaluation measures of classification have indicated that 13 features extracted by PCA method lead to better results than all of the extracted features. These results led to an accuracy of 82.2%, sensitivity of 77% and specificity of 86.93%. The proposed method may help dermatologists to detect the malignant lesions in the primary stages due to the minimum constraints during imaging, the ease of usage by the public and nonexperts, and high accuracy in detection of the lesion type.
Moallem, P. ,
Shokrani, S. ,
Habibi M. ,
Shokrani, S. ,
Moallem, P. ,
Habibi, M. 2025 29th International Computer Conference, Computer Society of Iran, CSICC 2025 pp. 215-220
Today Human Computer Interaction (HCI) is one of the most important topics in machine vision and image processing fields. Through features can get beneficial information about the variety of emotions and gestures which are produced by the movements of facial major parts. In this paper we presented the technique of Pyramid Histogram of Oriented Gradient for feature extraction and compare it with gabor filters. Six basic facial expressions plus the neutral pose are considered in the evaluations. The KNN and SVM techniques are used in the classification phase. Unlike most emotion detection approaches that focus on frontal face view this method concentrates on three views of the face and can easily be generalized to other poses and feelings. We have tested our algorithm on the Radboud faces database (RaFD) over three directions of head (frontal, 45 degree to the right and 45 degree to the left). Cohn-Kanade (CK+) and JAFFE are two other databases used in this work. The experiments using the proposed method demonstrate favorable results. In the best condition by using Pyramid Histogram Of Oriented Gradient plus KNN classification, the success rates were 100, 96.7, 98.1, 98.3 and 98.9 % for RaFD (frontal pose), RaFD (45 degree to the right), RaFD (45 degree to the left), JAFFE and CK+ databases respectively. © 2014 IEEE.
INTERNATIONAL JOURNAL OF ADVANCED ROBOTIC SYSTEMS (17298814) 11
In this paper, we designed a fast and low-cost mechatronic system for recognition of eight current Persian banknotes in circulation. Firstly, we proposed a mechanical solution for avoiding extra processing time caused by detecting the place of banknote and paper angle correction in an input image. We also defined new parameters for feature extraction, including colour features (RGBR values), size features (LWR) and texture features (CRLVR value). Then, we used a Multi-Layer Perceptron (MLP) neural network in the recognition phase to reduce the necessary processing time. In this research, we collected a perfect database of Persian banknote images (about 4000 double-sided prevalent images). We reached about 99.06% accuracy (average for each side) in final banknote recognition by testing 800 different worn, torn and new banknotes which were not part of the initial learning phase. This accuracy could increase to 99.62% in double-sided decision mode. Finally, we designed an ATmega32 microcontroller-based hardware with 16MHz clock frequency for implementation of our proposed system which can recognize sample banknotes at about 480ms and 560ms for single-sided detection and double-sided detection respectively, after image scanning.
Moallem, P. ,
Malekzadeh, M. ,
Zadeh F.K. ,
Asiri, S. ,
Zadeh F.K. ,
Moallem, P. ,
Asiri, S. ,
Malekzadeh, M. 2025 29th International Computer Conference, Computer Society of Iran, CSICC 2025 pp. 890-895
This article presents the results of a study on a prototype of spherical rolling robot and proposed a linear quadratic regulator controller to stabilize the system. The dynamic model of this spherical rolling robot has been presented with 2-DOF pendulum located inside a spherical shell and considered as a plate-ball system. The motion of the system is generated with a servo motor for left and right direction and a DC motor for forward and backward motion. Dynamic equations of the system are derived based on Euler-Lagrange method. Controlling a spherical robot is a challenging problem till today due to its nature of kinematics and highly nonlinear dynamics. Accordingly, a mid loop linear quadratic regulator (LQR) has been designed using full-state feedback to control the spherical robot. Simulation and experimental test have been carried out to show the effectiveness of the proposed controller. © 2014 IEEE.
Moallem, P. ,
Ehsanpour, M. ,
Bolhasani, A. ,
Montazeri, M. Journal of Electronics (19930615) 31(5)pp. 394-405
Arithmetic Logic Unit (ALU) as one of the main parts of any computing hardware plays an important role in digital computers. In quantum computers which can be realized by reversible logics and circuits, reversible ALUs should be designed. In this paper, we proposed three different designs for reversible 1-bit ALUs using our proposed 3×3 and 4×4 reversible gates called MEB3 and MEB4 (Moallem Ehsanpour Bolhasani) gates, respectively. The first proposed reversible ALU consists of six logical operations. The second proposed ALU consists of eight operations, two arithmetic, and six logical operations. And finally, the third proposed ALU consists of sixteen operations, four arithmetic operations, and twelve logical operations. Our proposed ALUs can be used to construct efficient quantum computers in nanotechnology, because the proposed designs are better than the existing designs in terms of quantum cost, constant input, reversible gates used, hardware complexity, and functions generated. © 2014, Science Press, Institute of Electronics, CAS and Springer-Verlag Berlin Heidelberg.
Iranian Journal Of Fuzzy Systems (17350654) 11(6)pp. 47-65
Potato image segmentation is an important part of image-based potato defect detection. This paper presents a robust potato color image segmentation through a combination of a fuzzy rule based system, an image thresholding based on Genetic Algorithm (GA) optimization and morphological operators. The proposed potato color image segmentation is robust against variation of background, distance and view of potato from digital camera. In the proposed algorithm, after selecting appropriate color space, distance between an image pixel and real potato pixels is computed. Furthermore, this distance feeds to a fuzzy rule-based classifier to extract potato candidate in the input image. A subtractive clustering algorithm is also used to decide on the number of rules and membership functions of the fuzzy system. To improve the performance of the fuzzy rule-based classifier, the membership functions shapes are also optimized by the GA. To segment potatoes in the input color image, an image thresholding is applied to the output of the fuzzy system, where the corresponding threshold is optimized by the GA. To improve the segmentation results, a sequence of some morphological operators are also applied to the output of thresholding stage. The proposed algorithm is applied to different databases with different backgrounds, including USDA, CFIA, and obtained potato images database from Ardabil (Iran’s northwest), separately. The correct segmentation rate of the proposed algorithm is approximately 98% over totally more than 500 potato images. Finally, the results of the proposed segmentation algorithm are evaluated for some images taken from real environments of potato industries and farms. © 2014 University of Sistan and Baluchestan. All rights reseved.
International Journal of Innovative Computing, Information and Control (13494198) 9(3)pp. 939-953
To improve transparency, as an important parameter in watermarking, and maintain robustness, a new quantization coefficient of the third and fourth sub-bands of Discrete Wavelet Transform (DWT) is being proposed. In this method, all coefficients of four-level Haar DWT sub-bands of a host image (HL4, LH4, HL3 and LH3) are divided into different non-overlapping blocks. Then, each block is divided into some sets consisting of several wavelet coefficients as their members. Depending on whether a zero or one needs to be embedded, one or all sets are selected. By quantizing the first and second largest coefficient values in each selected set, a watermark bit is embedded. In decoding stage, the lowest difference between the first and second largest coefficient values in each block is compared with an empirical threshold, in order to estimate the watermark bit. In comparison with other methods, the implementation results show that the proposed method sigficantly improves transparency, while enhancing the robustness for most attacks. Moreover, the proposed method establishes a trade-off between transparency and robustness by tuning a threshold value in the decoding stage. © 2013 ICIC International.
International Journal of Robotics and Automation (19257090) 28(2)pp. 137-145
Detection of external defects on potatoes is one of the most important technologies in the realization of automatic potato grading stations. In this paper, a computer vision-based potato defect detection algorithm using artificial neural networks and support vector machine (SVM) is proposed. In this algorithm, to detect the potatoes pixels in background, the supervised colour segmentation based on multilayer perceptrons (MLPs), radial basis function (RBF) neural networks as well as SVM are applied to the RGB component of each pixel. Afterwards, co-occurrence texture features are extracted from the grey level component of colour-space image, and finally three different classifiers including MLP, RBF and SVM are trained and validated to apply for defect detection. Results showed that the SVM classifiers represent a higher performance than the MLP and RBF neural networks for potato defect detection. The computational cost of the proposed SVM-based algorithm shows the possibility of a real-time implementation.
International Journal of Advanced Robotic Systems (17298814) 10
Gender classification from face images has many applications and is thus an important research topic. This paper presents an approach to gender classification based on shape and texture information gathered to design a fuzzy decision making system. Beside face shape features, Zernik moments are applied as system inputs to improve the system output which is considered as the probability of being male face image. After parameters tuning of the proposed fuzzy decision making system, 85.05% classification rate on the FERET face database (including 1199 individuals from different poses and facial expressions) shows acceptable results compare to other methods. © 2013 Moallem and Mousavi.
Transactions of the Institute of Measurement and Control (14770369) 35(3)pp. 342-352
The IEC 61000-4-15 flickermeter has been widely accepted as an international standard for flicker severity measurement. In this paper, the implementation of an improved IEC flickermeter by an ARM microcontroller-based digital system with low cost and simple proposed prototype hardware is presented. The improvements include replacing a band-pass filter with a more appropriate filter and considering the model of various lamps. The operation of the improved flickermeter is evaluated in a simulated power system in Simulink of MATLAB software with its flicker source an arc furnace. Finally, a manipulated hardware for implementation of the improved flickermeter is proposed based on ARM microcontroller, and its performance is evaluated with some real-time experimental measurements. © The Author(s) 2012.
Ashourian m., M. ,
Daneshmandpour n., ,
Sharifi-tehrani, O. ,
Moallem, P. International Journal of Engineering, Transactions B: Applications (1728144X) 26(11)pp. 1347-1356
License plate recognition (LPR) using morphology has the advantage of higher resistance to changes of brightness, high speed processing, and low complexity. However, these approaches are sensitive to the distance of the plate from the camera and imaging angle. Various assumptions reported in other works might be unrealistic, and cause major problems in practical experiences. In this paper we considered morphological approaches and improved them using adaptive techniques in order to provide more compatibility with practical applications. We examined the developed system on several car plate image databases with different conditions such as different camera distance, and different car views. The average achieved rate of success was 89.95% for all car plate location recognition, which is more than 6.0% improvement in comparison to previous morphological methods. We further developed and implemented an FPGA realization of the pre-processing stage of the system which is the main computation load of our LPR system.
Journal of the Chinese Institute of Engineers, Transactions of the Chinese Institute of Engineers,Series A/Chung-kuo Kung Ch'eng Hsuch K'an (02533839) 35(4)pp. 421-429
The flicker phenomenon, as a power quality aspect, can be measured by the standard IEC flickermeter that has been widely accepted as an international standard for flicker severity measurement. The standard IEC flickermeter is unable to measure the flicker effect of various lamps. It is possible that interharmonics appear in a system that causes flicker in some kinds of lamps but the IEC flickermeter shows that flicker does not exist in those systems. In this paper, this deficiency is evaluated and an appropriate way for considering the effect of various lamps is presented by decomposing the lamp-eye-brain model into two parts. From laboratory work, the gain factors of two kinds of lamps are obtained and evaluated using the measuring system of the IEC flickermeter. In order to estimate the gain factor of a lamp through an appropriate transfer function, the particle swarm optimization (PSO) algorithm is also applied. In a simulated system whose flicker source is a welding system, flicker effect on various lamps is evaluated. © 2012 The Chinese Institute of Engineers.
Conference Proceedings - IEEE International Conference on Systems, Man and Cybernetics (1062922X) pp. 2173-2177
This paper describes an automated algorithm for shadow region detection in Intra Vascular Ultrasound images, using an adaptive threshold method for threshold selection and contour approach for border detection. As shadow appears behind the calcification plaque, it makes it difficult or impossible for the dark region to process automatically aroundf these regions. The acoustic shadow usually follows the hard plaque in IVUS images and it can distinguish calcification regions from other bright regions. Therefore we propose to use Otsu Threshold for calcification plaque segmentation and the Active contours without edge method for shadow region separation of the image. Results show that the proposed meth efficiently detected shadow regions even in complicated images. This proposed algorithm presented specificity of 86% and sensitivity of 93%. © 2012 IEEE.
Eurasip Journal on Advances in Signal Processing (16876172) 2012(1)
One of the popular approaches in object boundary detecting and tracking is active contour models (ACM). This article presents a new balloon energy in parametric active contour for tracking a texture object in texture background. In this proposed method, by adding the balloon energy to the energy function of the parametric ACM, a precise detection and tracking of texture target in texture background has been elaborated. In this method, texture feature of contour and object points have been calculated using directional Walsh-Hadamard transform, which is a modified version of the Walsh-Hadamard. Then, by comparing the texture feature of contour points with texture feature of the target object, movement direction of the balloon has been determined, whereupon contour curves are expanded or shrunk in order to adapt to the target boundaries. The tracking process is iterated to the last frames. The comparison between our method and the active contour method based on the moment demonstrates that our method is more effective in tracking object boundary edges used for video streams with a changing background. Consequently, the tracking precision of our method is higher; in addition, it converges more rapidly due to it slower complexity. © 2012 Tahvilian et al.; licensee Springer.
Journal of Applied Research and Technology (16656423) 10(5)pp. 703-712
Selecting an optimal threshold value is the most important step in image thresholding algorithms. For a bimodal histogram which can be modeled as a mixture of two Gaussian density functions, estimating these densities in practice is not simply feasible. The objective of this paper is to use adaptive particle swarm optimization (APSO) for the suboptimal estimation of the means and variances of these two Gaussian density functions; then, the computation of the optimal threshold value is straightforward. The comparisons of experimental results in a wide range of complex bimodal images show that this proposed thresholding algorithm presents higher correct detection rate of object and background in comparison to the other methods including Otsu's method and estimating the parameters of Gaussian density functions using genetic algorithm (GA). Meanwhile, the proposed thresholding method needs lower execution time than the PSO-based method, while it shows a little higher correct detection rate of object and background, with lower false acceptance rate and false rejection rate.
Analog Integrated Circuits and Signal Processing (09251030) 71(2)pp. 337-341
JPEG coder is still the most attractive compression scheme for still image. Using Huffman coding makes the JPEG coder very sensitive to communication error. In this paper we propose a simple modification to JPEG coder by using reversible variable length code designed for the DC coefficients of each block. The results show great improvement in error resilience in a binary symmetric channel transmission with an average reduction of 5-10% in compression ratio.
This paper is concerned with the control of three-phase Uninterruptible Power Supply (UPS) systems based on a B-spline Network (BSN). A UPS must be able to maintain regulated output voltage in the face of all line/load disturbances. To achieve this, the control system of a UPS must have a fast transient response and a low steady-state error. Fast controllers cannot always maintain the required steady-state accuracy. Therefore, in this paper, a hybrid control solution is proposed. In the proposed method, a fast yet simple controller based on deadbeat (DB) control law is used to achieve the fast transient response required for UPS output voltage control. A B-spline based controller is added to the deadbeat controller to improve steady-state performance of the UPS system. This results in distortion-free output voltage along with a fast error convergence. The system modeling and controller design for the proposed structure are presented in this paper. Simulations results are shown for verification of the theoretical analysis. © 2011 IEEE.
Moallem, P. ,
Mousavi b.s., B.S. ,
Monadjemi, A. ,
Moallem, P. ,
Mousavi b.s., B.S. ,
Monadjemi, A. Applied Soft Computing (15684946) 11(2)pp. 1801-1810
Human face detection plays an important role in a wide range of applications such as face recognition, surveillance systems, video tracking applications, and image database management. In this paper, a novel fuzzy rule-based system for pose, size, and position independent face detection in color images is proposed. Subtractive clustering method is also applied to decide on the numbers of membership functions. In the proposed system, skin-color, lips position, face shape information and ear texture properties are the key parameters fed to the fuzzy rule-based classifier to extract face candidate in an image. Furthermore, the applied threshold on the face candidates is optimized by genetic algorithm. The proposed system consists of two main stages: the frontal/near frontal face detections and the profile face detection. In the first stage, skin and lips regions are identified in HSI color space, using fuzzy schemes, where the distances of each pixel color to skin-color and lips-color clusters are applied as the input and skin-likelihood and lips-like images are produced as the output. Then, the labeled skin and lips regions are presented to both frontal and profile face detection algorithms. A fuzzy rule-based containing the face and lips position data, along with the lips area and face shape are employed to extract the frontal/near frontal face regions. On the other hand, the profile face detection algorithm uses a geometric moments-based ear texture classification to verify its outcomes. The proposed method is tried on various databases, including HHI, Champion, Caltech, Bao, Essex and IMM databases. It shows about 98, 96 and 90% correct detection rates over 783 samples, in frontal, near frontal and profile face images, respectively. © 2010 Elsevier B.V. All rights reserved.
Optical Review (13406000) 18(6)pp. 415-422
Though a significant amount of work has been done on detecting obstacles, not much attention has been given to the detection of drop offs, e. g., sidewalk curbs, downward stairs, and other hazards. In this paper, we propose algorithms for detecting negative obstacles in an urban setting using stereo vision and two-stage dynamic programming (TSDP) technique. We are developing computer vision algorithms for sensing important terrain features as an aid to blind navigation, which interpret visual information obtained from images collected by cameras mounted on camera legs nearly as high as young person. This paper focuses specifically on a novel computer vision algorithm for detecting negative obstacles (i. e. anything below the level of the ground, such as holes and drop-offs), which are important and ubiquitous features on and near sidewalks and other walkways. The proposed algorithm is compared to other algorithms such as belief propagation and random growing correspondence seeds (GCS). According to the results, the proposed method achieves higher speed, more accurate disparity map and lower RMS errors. The speed of the proposed algorithm is about 28% higher than the random GCS algorithm. We demonstrate experimental results on typical sidewalk scenes to show the effectiveness of the proposed method. © 2011 The Japan Society of Applied Physics.
2025 29th International Computer Conference, Computer Society of Iran, CSICC 2025
Active contour model (ACM) is a powerful tool for the target tracking in digital image sequences. Traditional ACM fails to track the targets when target displacements or aspect changes are high and inhomogeneous. In order to improve these inefficiencies, in this paper, a modified balloon energy in the parametric ACM is defined base on target displacement in two successive frames to inflate some suitable points adaptively and locally, in the current frame. Moreover, traditional ACM suffers from low capture range, so it can not attract to the concave boundaries of the target those results to an uncertain tracking in image sequences. To improve this inefficiency, a modified virtual electric field energy is used in conjunction with the proposed balloon energy. Also, new ACM algorithm adapted base on mutual changes of target and background gray levels. The advantages of the proposed ACM consist of: less sensitivity to initialization, large capture range, attraction to sharp-pointed and concave boundaries of target, ability of tracking the target with large and inhomogeneous displacements, and acceptable computational cost. Experimental results show that the tracking by the proposed ACM based on the greedy algorithm produces more correct detection percent of the target boundaries than other similar methods, in different circumstances. © 2011 AmirKabir Univ of Tech.
2025 29th International Computer Conference, Computer Society of Iran, CSICC 2025 pp. 264-267
Today, Karyotype analysis is frquently used in cytogenetics. it is time-consuming and repetitive, so an automatic analysis can greatly be valued. In this reasearch, an automatic method is presented. it first uses the geometric shape of the clusters to devide them into single chromosomes and clusters with more than one chromosome. Then each cluster is investigated to find the dark paths in touching chromosomes and geometric shapes to separate overlapping ones. A criterion function decides about similarity between the outputs and the single chromosomes and the best one is chosen. precision of this method in recognition of clusters with more than one chromosome is 75% and in separation of touching and overlapping ones are 74% and 66% respectively. © 2011 IEEE.
Journal of Power Electronics (15982092) 11(2)pp. 228-236
The δV10 or 10-Hz flicker index, as a common method of measurement of voltage flicker severity in power systems, requires a high computational cost and a large amount of memory. In this paper, for measuring the δV10 index, a new method based on the Adaline (adaptive linear neuron) system, the FFT (fast Fourier transform), and the PSO (particle swarm optimization) algorithm is proposed. In this method, for reducing the sampling frequency, calculations are carried out on the envelope of a power system voltage that contains a flicker component. Extracting the envelope of the voltage is implemented by the Adaline system. In addition, in order to increase the accuracy in computing the flicker components, the PSO algorithm is used for reducing the spectral leakage error in the FFT calculations. Therefore, the proposed method has a lower computational cost in FFT computation due to the use of a smaller sampling window. It also requires less memory since it uses the envelope of the power system voltage. Moreover, it shows more accuracy because the PSO algorithm is used in the determination of the flicker frequency and the corresponding amplitude. The sensitivity of the proposed method with respect to the main frequency drift is very low. The proposed algorithm is evaluated by simulations. The validity of the simulations is proven by the implementation of the algorithm with an ARM microcontroller-based digital system. Finally, its function is evaluated with real-time measurements.
Australian Journal of Basic and Applied Sciences (19918178) 5(11)pp. 2040-2045
Discrete cosine transform (DCT) is the fundamental part of JPEG compressor and is one of the most widely used conversion technique in digital signal processing (DSP) and image compression. Due to importance of the discrete cosine transform in JPEG standard, an algorithm is proposed that is in parallel structure thus intensify hardware implementation speed of discrete cosine transform and JPEG compression procedure. The proposed method is implemented by utilizing VHDL hardware description language in structural format and follows optimal programming tips by which, low hardware resource utilization, low latency, high throughput and high clock rate are achieved. Inputs are 8-bit long, 4 separate units are considered and CSA and CLA adders are used to realize discrete cosine transform. Working frequency for this implementation is 100 MHz and each stage delay is 10ns which is optimum in comparison with other methods. This proposed method can be easily utilized in any hardware applications such as JPEG compressor, image/signal processing and etc. by minimum change in design parameters. Also, it can be used as a hard-core in embedded systems, system on chips (SOC), system on programmable chips (SOPC) and network on chips (NOC).
International Journal of Computer Mathematics (00207160) 88(1)pp. 21-36
As the derivative of the sigmoid activation function approaches zero, the back propagation algorithm involves flat spots on the error surface of multilayer perceptron (MLP) neural networks, which means the hidden neurons of MLP were saturated. Flat spots can slow down the gradient search and hamper convergence. In this paper, we propose a grading technique to gradually level off the potential flat spots to a sloping surface in a look-ahead mode; and thereby progressively renew the saturated hidden neurons. We introduce a criterion to measure the saturation level of MLP, and then we modify the error function by using a proposed piecewise error function that switches between two cases, regarding the level of MLP saturation. These two cases include the standard error function, when MLP is not saturated, and the modified error function, when MLP is saturated. We recorded considerable improvements, especially in convergence rate and generalization, on the tested benchmark problems. © 2011 Taylor & Francis.
Iranian Journal of Electrical and Computer Engineering (16820053) 9(2)pp. 127-133
In feed forward neural networks, hidden layer neurons' saturation conditions, which are the cause of flat spots on the error surface, is one of the main disadvantages of any conventional gradient descent learning algorithm. In this paper, we propose a novel complementary scheme for the learning based on a suitable combination of anti saturated hidden neurons learning process and accelerating methods like the momentum term and the parallel tangent technique. In our proposed method, a normalized saturation criterion (NSC) of hidden neurons, which is introduced in this paper, is monitored during learning process. When the NSC is higher than a specified threshold, it means that the algorithm moves towards a flat spot as the hidden neurons fall into saturation condition. In this case, in order to suppress the saturation of hidden neurons, a conventional gradient descent learning method can be accompanied by the proposed complementary gradient descent saturation prevention scheme. When the NSC assumes small values, no saturation detected and the network operates in its normal condition. Therefore, application of a saturation prevention scheme is not recommended. We have evaluated the proposed complementary method in accompaniment to the gradient descent plus momentum and parallel tangent, two conventional improvements on learning methods. We have recorded remarkable improvements in convergence success as well as generalization in some well known benchmarks. © 2010.
2025 29th International Computer Conference, Computer Society of Iran, CSICC 2025 pp. 194-199
Stereo vision based obstacle detection is an algorithm that aims to detect and compute obstacle depth using stereo matching and disparity map. This paper presents a novel method to detect obstacles in highly textured environments using two-stage dynamic programming. The algorithm consists of several steps including pre-processing, obstacle detection, analysis of disparity map using two-stage dynamic programming (TSDP) technique and depth computation. This method works well in highly textured environments and ideal for real applications. The disparity map for the stereo images is found in the 3D correlation coefficient volume by obtaining the global 3D maximum-surface rather than simply choosing the position that gives the local maximum correlation coefficient value for each pixel. The 3D maximum-surface is obtained using two-stage dynamic programming (TSDP) technique. An adaptive thresholding is also applied for better noise and texture removal. Experimental results show the effectiveness of the proposed method. ©2010 IEEE.
International Journal of Engineering, Transactions B: Applications (1728144X) 23(2)pp. 121-132
Motion-JPEG is a common video format for compression of motion images with high quality using JPEG standard for each frame of the video. During transmission through a noisy channel some blocks of data are lost or corrupted, and the quality of decompression frames decreased. In this paper, for reconstruction of these blocks, several temporal-domain, spatial-domain, and frequency-domain error concealment methods are investigated. Then a novel method is proposed for recovery of channel errors with a mixture of temporal-domain and frequency-domain error concealment methods. To reconstruct the missed blocks in the proposed novel method, when two successive frames are similar, a proposed two phase block matching algorithm is performed in temporal-domain. When two successive frames are different, our proposed method reconstructs the missed block by the estimation of DC and AC coefficient, in frequency-domain. The proposed method and the other similar methods are simulated for different noise and quality factors. The results of quality measurements are indicated that in all tested video sequences, the proposed method shows higher quality in reconstruction of missed blocks.
World Academy of Science, Engineering and Technology (20103778) 43pp. 699-704
This paper presents a robust method to detect obstacles in stereo images using shadow removal technique and color information. Stereo vision based obstacle detection is an algorithm that aims to detect and compute obstacle depth using stereo matching and disparity map. The proposed advanced method is divided into three phases, the first phase is detecting obstacles and removing shadows, the second one is matching and the last phase is depth computing. We propose a robust method for detecting obstacles in stereo images using a shadow removal technique based on color information in HIS space, at the first phase. In this paper we use Normalized Cross Correlation (NCC) function matching with a 5 × 5 window and prepare an empty matching table τ and start growing disparity components by drawing a seed s from S which is computed using canny edge detector, and adding it to τ. In this way we achieve higher performance than the previous works [2,17]. A fast stereo matching algorithm is proposed that visits only a small fraction of disparity space in order to find a semi-dense disparity map. It works by growing from a small set of correspondence seeds. The obstacle identified in phase one which appears in the disparity map of phase two enters to the third phase of depth computing. Finally, experimental results are presented to show the effectiveness of the proposed method.
Connection Science (13600494) 22(4)pp. 373-392
Poor convergence is a common problem of gradient-based multi-layer perceptron (MLP)-learning algorithms. It is claimed that using a deflecting direction like momentum and adaptive learning rates (ALRs) can improve the convergence performance. For a more reliable and faster MLP learning, we introduce the parallel tangent gradient with adaptive learning rates (PTGALR) algorithm that uses parallel tangent deflecting direction instead of the momentum. Moreover, we use two independent variable learning rates, one for the gradient descent and the other for accelerating direction through the parallel tangent. Also, we propose an improved ALR computation algorithm that calculates the learning rates with a dispensable error oscillation. This adaptation algorithm has two outputs: one is the applied learning rate and the other is used for a better learning rate estimation in the next iteration. Moreover, the proposed ALR computation algorithm models the error function as a one-dimensional quadratic function of the learning rate when itis needed. The implementation results of PTGALR for some well-known binary and real MLP problems show higher and faster convergence with lower oscillations than the similar adaptive learning algorithms. © 2010 Taylor & Francis.
International Review on Computers and Software (discontinued) (18286003) 5(4)pp. 436-443
An FPGA-based fixed-point standard-LMS algorithm core is proposed for adaptive signal processing (ASP) realization in real time. The LMS core is designed in VHDL93 language as basis of FIR adaptive filter. FIR adaptive filters are mostly used because of their low computation costs and linear phase. The proposed model uses 12-bit word-length for input data from analog to digital converter (ADC) chip while internal computations are based on 17-bit word-length because of considering guard bits to prevent overflow. The designed core is FPGAbrand- independent so that it can be implemented on any brand to create a system-onprogrammable- chip (SoPC). In this paper, XILINX SPARTAN3E and VIRTEX4 FPGA series are used as implementation platform. Rounding errors were inevitable due to limited word-length and can be decreased by adjusting the dynamic range of input signal amplitude. A comparison is made between DSP, Hardware/Software co-design and pure-hardware implementations. Obtained results show improvements in area-resource utilization, convergence speed and performance in the designed pure hardware LMS core. Although using a pure-hardware implementation results in high performance, it is much more complex than other structures. © 2010 Praise Worthy Prize S.r.l.
Moallem, P. ,
Koleini, M. ,
Monadjemi, A. ,
Koleini, M. ,
Monadjemi, A. ,
Moallem, P. Journal of the Chinese Institute of Engineers, Transactions of the Chinese Institute of Engineers,Series A (2533839) 33(7)pp. 1049-1057
This study represents an innovative automatic method for black and white films colorization using texture features and a multilayer perceptron artificial neural network. In the proposed method, efforts are made to remove human interference in the process of colorization and replace it with an artificial neural network (ANN) which is trained using the features of the reference frame. Later, this network is employed for automatic colorization of the remained black and white frames. The reference frames of the black and white film are manually colored. Using a Gabor filter bank, texture features of all the pixels of the reference frame are extracted and used as the input feature vector of the ANN, while the output will be the color vector of the corresponding pixel. Finally, the next frames’ feature vectors are fed respectively to the trained neural network, and color vectors of those frames are the output. Applying AVI videos and using various color spaces, a series of experiments are conducted to evaluate the proposed colorization process. This method needs considerable time to provide a reasonable output, given rapidly changing scenes. Fortunately however, due to the high correlation between consecutive frames in typical video footage, the overall performance is promising regarding both visual appearance and the calculated MSE error. Apart from the application, we also aim to show the importance of the low level features in a mainly high level process, and the mapping ability of a neural network. © 2010, Taylor & Francis Group, LLC.
Multiplier circuits play an important role in reversible computation, which is helpful in diverse areas such as low power CMOS design, optical computing, DNA computing and bioinformatics, quantum computing and nanotechnology. In this paper a new reversible device called MFA (modified full adder) is used to design a novel reversible 4-bit binary multiplier circuit with low hardware complexity. It has been shown that the proposed reversible logic device in designing multiplier circuits can work singly as a reversible full adder. Furthermore, it has been demonstrated that the proposed design of reversible multiplier circuit needs fewer garbage outputs and constant inputs. The proposed multiplier can be generalized for NxN bit multiplication. Thus, this job will be of significant value as the technologies mature. © 2010 IEEE.
Journal of Computers (discontinued) (1796203X) 5(7)pp. 1094-1099
In this study a new artificial neural network based approach to automatic or semi-automatic colorization of black and white film footages is introduced. Different features of black and white images are tried as the input of a MLP neural network which has been trained to colorize the movie using its first frame as the ground truth. Amongst the features tried, e.g. position, relaxed position, luminance, and so on, we are most interested on the texture features namely the Laws filter responses, and what their performance would be in the process of colorization. Also, the network parameter optimization, the effects of color reduction, and relaxed x-y position of pixels as the feature, are investigated in this study. The results are promising and show that the combination of MLP and texture features is effective in this application. © 2010 ACADEMY PUBLISHER.
Moallem, P. ,
Sharifi-tehrani, O. ,
Ashourian m., M. ,
Sharifi-tehrani, O. ,
Ashourian m., M. ,
Moallem, P. 2025 29th International Computer Conference, Computer Society of Iran, CSICC 2025
An FPGA-based channel noise canceller using a fixed-point standard-LMS algorithm for image transmission is proposed. The proposed core is designed in VHDL93 language as basis of FIR adaptive filter. The proposed model uses 12-bits word-length for digital input data while internal computations are based on 17-bits word-length because of considering guard bits to prevent overflow. The designed core is FPGA-brand-independent, thus can be implemented on any brand to create a system-on-programmable-chip (SoPC). In this paper, XILINX SPARTAN3E and VIRTEX4 FPGA series are used as implementation platform. A discussion is made on DSP, Hardware/Software co-design and pure-hardware implementations. Although using a pure-hardware implementation results in better performance, it is more complex than other structures. Results obtained show improvements in area-resource utilization, convergence speed and performance in the designed pure-hardware channel noise canceller core. © 2010 IEEE.
Neural Network World (23364335) 20(2)pp. 207-222
The gradient descent backpropagation (BP) algorithm that is widely used for training MLP neural networks can retard convergence due to certain features of the error surface like the local minimum and the flat spot. Common promoting methods, such as applying momentum term and using dynamic adaptation of learning rates, can enhance the performance of BP. However, saturation state of hidden layer neurons, which is the cause of some flat spots on the error surface, persists through such accelerating methods. In this paper, we propose a grading technique to gradually level off the potential flat spots into a sloping surface in a look-ahead mode; and thereby progressively renew saturated hidden neurons. We introduce symptoms indicating saturation state of hidden nodes. In order to suppress the saturation, we added a modifying term to the error function only when saturation is detected. In normal conditions, the improvement made to the learning process is adding a momentum term to the weight correction formula. We have recorded remarkable improvements in a selection of experiments. ©ICS AS CR 2010.
2025 29th International Computer Conference, Computer Society of Iran, CSICC 2025 pp. 177-182
A human face detection method for color images is presented in this paper. The system is composed of three subsystems: skin color segmentation, lip color segmentation and face blobs selection subsystem. Whole these algorithms are fuzzy rule base ones, which are designed empirically, and then optimized by genetic algorithm. In the first stage, skin color regions are selected in the input image. Within each of the skin area, lip pixels are searched using second subsystem, and applied as a feature to identify face candidates in the skin regions. Utilizing the lip area and position relative to the skin area, and face shape information, the third subsystem is materialized to choose face blobs. To precise evaluation of the proposed system, the false positive and false negative of each subsystem, are reported for the empirically designed system as well as the optimized system. Obtained results show a remarkable decrease in false positive and false negative for optimized algorithms compared to empirically designed ones. Finally, 98% detection rate is achieved using proposed method. ©2010 IEEE.
2025 29th International Computer Conference, Computer Society of Iran, CSICC 2025 pp. 944-949
In this paper we propose two learning algorithms for a spiking neural network which encodes information in the timing of spike trains. These algorithms are based on dynamic self adaptation for adapting the gradient learning rates (DS-η) and dynamic self adaptation for adapting the gradient learning rates and momentum (DS-ηα) algorithms. In our proposed algorithm, the optimum value for η was obtained from a parabolic function of error in both of these two algorithms and optimum value for α was obtained from our proposed adaptive algorithm. We performed a selection of benchmark problems to investigate the efficiency of our proposed algorithm. Compared to previously proposed algorithms such as SpikeProp and DS-ηα our algorithms, mod-DS-η and mod-DS-ηα, are faster than other methods in learning of the spiking neural networks. © 2010 IEEE.
International Review on Modelling and Simulations (19749821) 3(6)pp. 1241-1248
State space analysis can be successfully used to analyze the amplifier, load and the feed back circuits when they are all connected to each other. The main advantages of state space analysis of feedback amplifiers are the possibility of discussion of internal and structural nonlinearities and eigenvalue analysis of the whole amplifier. In this paper, state space analysis of a microwave amplifier is completely discussed. In this model, the effects of internal noise both in gate to source and drain to source junctions, are briefly taken into account. The transfer functions as well as input and output driving impedances of the amplifier are determined using proper state variables. This analysis will be valid when the network is in a linear region. © 2010 Praise Worthy Prize S.r.l. - All rights reserved.
Flicker phenomenon, as a power quality component, can be measured by the standard IEC flickermeter that has been widely accepted as an international standard for flicker severity measurement. In industry applications, the standard IEC flickermeter suffers from some deficiencies. Among them, it is unable to measure flickers caused by interharmonics whose frequencies are higher than 85Hz. It also is unable to measure the effect of the various lamp flickers. In addition, it is unable to determine direction of flicker source with respect to a monitoring point. In this paper, these deficiencies are improved respectively by designing an improved filter, using the border graph of gain factor of various lamps and by using flicker power to determine direction of a flicker source. At the end, the proposed flickermeter is evaluated in a simulated system, that its flicker source is a welding system. ©2010 IEEE.
Kuwait Journal Of Science And Engineering, Kuwait University (10248684) 36(2 B)pp. 91-105
Welding radiography is a common method in the quality control of welds. Good human interpreting of radiography images is time consuming. Therefore using intelligent methods in radiography interpreting can improve the quality of interpretations as well as reducing interpreting time. In X-Ray weld inspection systems with video output, it is necessary to detect defects very fast. As an assistant for weld images interpreters, this paper introduces an image processing algorithm with low computational cost for highlighting of cracks that may be lost in a video based X-Ray weld inspection system. The interpreter can then detect true cracks in the highlighted candidate positions faster and more accurate. To alleviate the noise effects in the digital input images, a 7 × 7 Gaussian filter with variance of 1.5 is applied. Then the vertical projections of the filtered images are used to accurately detect weld region of the input images. The detected weld region is segmented to rectangular regions. The horizontal projection of the rectangular regions is analyzed to highlight the transverse cracks, while the vertical projection of rectangular regions is used to highlight longitude cracks. Some parameters of the proposed algorithm can be adjusted by user to obtain required accuracy. The implementation results of the proposed algorithm for the real radiography weld images show the power of the algorithm to highlight transverse and longitude cracks. To evaluate the performance of the proposed image processing algorithm for video images in real time application, a visual C + + programming implementation of the algorithm is developed. The execution time of the developed application for video sequences in a PC with a 2.0 GHz Core Dou CPU is less than 40 milliseconds per image. Therefore it is suitable for real time cracks highlighting of online welding X-Ray inspection systems with video output.
Stereo vision based obstacle detection is an algorithm that aims to detect and compute obstacle depth using stereo matching and disparity map. This paper presents a robust method to detect positive obstacles including staircases in highly textured environments. The proposed method is easy to implement and fast enough for obstacle avoidance. This work is partly inspired by the work of Nicholas Molton et al [1]. The algorithm consists of several steps including calibration, pre processing, obstacle detection, analysis of disparity map and depth computation. This method works well in highly textured environments and ideal for real applications. An adaptive thresholding is also applied for better noise and texture removal. Experimental results show the effectiveness of the proposed method. © 2010 IEEE.
International Review of Electrical Engineering (25332244) 4(2)pp. 242-248
For determination the number of broken rotor bars in squirrel-cage induction motors when these motors are working, this paper presents a new method based on an intelligent processing of the stator transient starting current. In light load condition, distinguishing between safe and faulty rotors is difficult, because the characteristic frequencies of rotor with broken bars are very close to the fundamental component and their amplitudes are small in comparison. To overcome this problem, an advanced technique based on the wavelet transform and artificial neural network is suggested for processing the starting current of induction motors. In order to increase the efficiency of the proposed method, the results of the wavelet analysis, before applying to the neural networks are processed by Principal Component Analysis (PCA). Then the outcome results are supposed as neural network's training and testing data set. The trained neural networks undertake of determining the number of broken rotor bars. The given statistical results, announce the proposed method's high ability to determine the number of broken rotor bars. The proposed method is independent from loading conditions of machine and it is useable even when the motor is unloaded. © 2009 Praise Worthy Prize S.r.l. - All rights reserved.
Lecture Notes in Electrical Engineering (18761119) 27(VOL.1)pp. 105-111
The reconstruction of a dynamic complex 3D scene from multiple images is a fundamental problem in the field of computer vision. Given a set of images of a 3D scene, in order to recover the lost third dimension, depth, it is necessary to extract the relationship between images through their correspondence. Reduction of the search region in stereo correspondence can increase the performances of the matching process in both execution time and accuracy. In this study we employ edge-based stereo matching and hierarchical multiresolution techniques as fast and reliable methods, in which some matching constraints such as epipolar line, disparity limit, ordering, and limit of directional derivative of disparity are satisfied as well. The proposed algorithm has two stages: feature extraction and feature matching. We use color stereo images to increase the accuracy and link detected feature points into chains. Then the matching process is completed by comparing some of the feature points from different chains. We apply this new algorithm on some color stereo images and compare the results with those of gray level stereo images. The comparison suggests that the accuracy of our proposed method is increased around 20-55%. © 2009 Springer Science+Business Media, LLC.
Journal of Multimedia (discontinued) (17962048) 4(4)pp. 240-247
In this paper a novel method for machine-based black and white films colorization is presented. The kernel of the proposed scheme is a trained artificial neural network which maps the frame pixels from a grayscale space into a color space. We employ the texture coding method to capture the line/texture characteristics of each pixel as its most significant gray scale space feature, and using that feature, expect a highly accurate B/W to color mapping from the ANN. The ANN would be trained by the B/W-color pairs of an original reference frame. The experiments performed on some typical video footages show the advantages of the proposed method in both visual and mathematical aspects. Different color spaces are also tried to obtain the optimum colorization performance. © 2009 ACADEMY PUBLISHER.
International Journal of Imaging Systems and Technology (10981098) 19(3)pp. 179-186
In recent years, active contour models (ACM) have been considered as powerful tools for image segmentation and object tracking in computer vision and image processing applications. This article presents a new tracking method based on parametric active contour models. In the proposed method, a new pressure energy called "texture pressure energy" is added to the energy function of the parametric active contour model to detect and track a texture target object in a texture background. In this scheme, the texture features of the contour are calculated by a moment-based method. Then, by comparing these features with texture features of the target object, the contour curve is expanded or contracted to be adapted to the object boundaries. Experimental results show that the proposed method is more efficient and accurate in the tracking of objects compare to the traditional ones, when both object and background are textures in nature. © 2009 Wiley Periodicals, Inc.
Journal of Computer Science (15493636) 4(8)pp. 663-667
Problem Statement: Lung disease is a major threat to the human health regarding the industrial life, air pollution, smoking, and infections. Lung function tests are often performed using spirometry. Approach: The present study aims at detecting obstructive and restrictive pulmonary abnormalities. Lung function tests are often performed using spirometry. In this study, the data were obtained from 250 volunteers with standard recording protocol in order to detect and classify pulmonary diseases into normal, obstructive and restrictive. Firstly, spirometric data was statistically analyzed concerning its significance for neural networks. Then, such parameters were presented as input to MLP and recurrent networks. Results: These two networks detected normal and abnormal disorders as well as obstructive and restrictive patterns, respectively. Moreover, the output data was confirmed by measuring, accuracy and sensitivity. Conclusion: The results show that the proposed method could be useful for detecting the function of respiratory system. © 2008 Science Publications.
Journal of the Chinese Institute of Engineers, Transactions of the Chinese Institute of Engineers,Series A/Chung-kuo Kung Ch'eng Hsuch K'an (02533839) 31(4)pp. 649-657
In recent years, Active Contour Models (ACMs) have become powerful tools for object detection and image segmentation in computer vision and image processing applications. This paper presents a new energy function in parametric active contour models for object detection and image segmentation. In the proposed method, a new pressure energy called “texture pressure energy” is added to the energy function of the parametric active contour model to detect and segment a textured object against a textured background. In this scheme, the texture features of the contour are calculated by a moment based method. Then by comparing these features with texture features of the object, the contour curve is expanded or contracted in order to be adapted to the object boundaries. Experimental results show that the proposed method has more efficient and accurate segmenting functionality than the traditional method when both object and background have texture properties. © 2008, Taylor & Francis Group, LLC.
2025 29th International Computer Conference, Computer Society of Iran, CSICC 2025 pp. 55-60
Saturation conditions of the hidden layer neurons are a major cause of learning retardation in Multi Layer Perceptrons (MLP). Under such conditions the traditional backpropagation (BP) algorithm is trapped in local minima. To renew the search for a global minimum, we need to detect the traps and an offset scheme to avoid them. We have discovered that the gradient norm drops to a very low value in local minima. Here, adding a modifying term to the standard error function enables the algorithm to escape the local minima. In this paper, we proposed a piecewise error function; i.e. where the gradient norm remained higher than a parameter we used the standard error function, and added a modifying term to the function below this value. To further enhance this algorithm, we used our proposed adaptive learning rate schema. We performed a selection of benchmark problems to asses the efficiency of our proposed algorithm. Compared to previously proposed algorithms, we recorded higher convergence rates, especially in complex problems with complex input-output mapping. ©2008 IEEE.
2025 29th International Computer Conference, Computer Society of Iran, CSICC 2025 pp. 2162-2165
In this paper, we propose a MLP learning algorithm based on the parallel tangent gradient with modified variable learning rates, PTGVLR. Parallel tangent gradient uses parallel tangent deflecting direction instead of the momentum. Moreover, we use two separate and variable learning rates one for the gradient descent and the other for accelerating direction through parallel tangent. We test PTGVLR optimization method for optimizing a two dimensional Rosenbrock function and for learning of some well-known MLP problems, such as the parity generators and the encoders. Our investigations show that the proposed MLP learning algorithm, PTGVLR, is faster than similar adaptive learning methods. © ICROS.
Journal of Circuits, Systems and Computers (17936454) 16(2)pp. 305-317
Success of a tracking method depends largely on choosing the suitable window size as soon as the target size changes in image sequences. To achieve this goal, we propose a fast tracking algorithm based on adaptively adjusting tracking window. Firstly, tracking window is divided into four edge subwindows, and a background subwindow around it. Then, by calculating the spatiotemporal gradient power ratios of the target in each subwindow, four proper expansion vectors are associated with any tracking window sides such that the occupancy rate of the target in tracking window should be maintained within a specified range. In addition, since temporal changing of target is evaluated in calculating these vectors, we estimate overall target displacement by sum of expansion vectors. Experimental results using various real video sequences show that the proposed algorithm successfully track an unknown textured target in real time, and is robust to dynamic occlusions in complex noisy backgrounds. © World Scientific Publishing Company.
WSEAS Transactions on Computers (discontinued) (11092750) 5(3)pp. 469-476
Computations in a feature based stereo matching which is basically used for depth extraction are generally very high. These computations essentially include feature extraction and matching which feature matching is usually higher. For a feature-based stereo matching, we accurately tune the search space based on some stereo imaging parameters like the focal length with pixels scale, the displacement of features points and maximum disparity. We show that results of previous matches can be used to narrow down the search space to find current match. We use directional derivative of disparity as a temporary concept to tune the search space accurately. Then we develop a fast feature based stereo matching algorithm based on the proposed search space tuning and non-horizontal thinned edge points as features. For reducing the error in the matching stage, we use left-right consistency checking technique for a small number of feature points. Usually this technique doubles the execution time of matching, but we will show that increasing of execution time in the proposed algorithm is negligible respect to the other similar methods as well as the invalid matching is also highly reduced. Comparing to the other similar methods, experimental results for three tested images show that not only the execution time of the matching stage of the proposed algorithm is decreased to 42%, but also the error in the matching is decreased to 90%.
WSEAS Transactions on Computers (discontinued) (11092750) 5(9)pp. 2106-2113
A new Walsh/Hadamard-based texture feature extraction method would be proposed in this paper. This method, called Directional Walsh/Hadamard Transform or DWHT, tries to insert the basic advantages of multi-scale/ multi-directional texture analysis into a fast modified Walsh/Hadamard transform. A comparative test amongst several texture feature extraction algorithms applied on a database of VisTex. textures, illustrates the efficiency of the proposed method in both classification accuracy and the computational costs.
In this paper, an optimization method based on Genetic Algorithms (GA) is applied to find the best design parameters of the switching power circuit for a Switched Reluctance Motor (SRM). The optimal parameters are found by GA with two objective functions, i.e. efficiency and torque ripple. A fuzzy expert system for predicting the performance of a switched reluctance motor has been developed. The design vector consists of design parameters, and output performance variables are efficiency and torque ripple. An accurate analysis program based on Improved Magnetic Equivalent Circuit (IMEC) method has been used to generate the input-output data. These inputoutput data are used to produce the optimal fuzzy rules for predicting the performance of SRM. Table look-up scheme and gradient decent training are used for optimal fuzzy prediction designed. The results of the optimal switching power circuit design for a 8/6, four phase, 4 kW, 250V, 1500rpm SR motor © 2006 IEEE.
Moallem, P. ,
Dehkordi, B.M. ,
Monadjemi, A. ,
Ashourian m., M. WSEAS Transactions on Computers (discontinued) (11092750) 5(9)pp. 2098-2105
As a gradient based optimization algorithm, we introduce PATLIS (PArallel Tangent and heuristic LIne Search) optimization method that can be used as a learning algorithm for MLP neural networks. In typical gradient based learning algorithms, the momentum rate has usually an improving effect in convergence rate and decreasing the zigzagging phenomena. However it sometimes causes the convergence rate to decrease. The Parallel Tangent (ParTan) gradient can be used as deflecting method to improve the convergence. From the implementation point of view, it is as simple as the momentum. In fact this method is one of the more practical implementation of conjugate gradient. ParTan tries to overcome the inefficiency of zigzagging phenomena of conventional backpropagation by deflecting the gradient through acceleration phase. In this paper, we use two learning rates, η for gradient search direction and μ for accelerating direction through parallel tangent. Moreover, a heuristic line search based on an improved version of dynamic self adaptation of η and μ is used to improve the proposed learning method. In dynamic self adaptation, each learning rate is adapted locally to the cost function landscape and the previous learning rate. Finally we test the proposed algorithm for optimizing Rosenbrock function and for various MLP neural networks including a XOR 2×2×1, Encoder 16×4×16 and finally Parity 4×4×1. We compare the results with those of the dynamic self adaptation of gradient learning rate and momentum (DSη-α) and parallel tangent with dynamic self adaptation (PTDSη-μ). Experimental results for optimizing Rosenbrock function for the first 100 iterations of executions showed the convergence speed of PATLIS is very faster than the other tested methods. Furthermore for MLP problems, the experimental results showed that the average numbers of epochs for PATLIS respect to PTDSη-μ and DSη-α were decreased to around 50% and 66% respectively. Our proposed algorithm also shows a good power for avoiding from local minimum.
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (03029743) 3768pp. 258-269
In this paper, we propose a methodology for embedding watermark image data into color images. At the transmitter, the signature image is encoded by a multiple description encoder. The information of the two descriptions are embedded in the host image in the spatial domain in the red and blue components. Furthermore, this scheme requires no knowledge of the original image for the recovery of the signature image, yet yields high signal-to-noise ratios for the recovered output. At the receiver, the multiple description decoder combines the information of each description and reconstructs the original signature image. We experiment the proposed scheme for embedding a gray-scale signature image of 128x128 pixels size in the spatial domain of a color host image of 512x512 pixels. Simulation results show that data embedding based on multiple description coding has low visible distortions in the host image and robustness to various signal processing and geometrical attacks, such as addition of noise, quantization, cropping and down-sampling.
Journal of Circuits, Systems and Computers (17936454) 14(2)pp. 249-266
Reduction of search region in stereo correspondence can increase performances of the matching process, in the context of execution time and accuracy. For an edge-based stereo matching, we establish relationships between the search space and the parameters like relative displacement of edges, disparity under consideration, image resolution, CCD (Charge-Coupled Device) dimension and focal length of the stereo system. Then, we propose a novel matching strategy for the edge-based stereo. Afterward, we develop a fast edge-based stereo algorithm with combination of the obtained matching strategy and a multiresolution technique using Haar wavelet. Considering the conventional multiresolution technique using Haar wavelet, the execution times of our proposed method are decreased between 26% to 47% in the feature matching stage. Moreover, the execution time of the overall algorithms (including feature extraction and feature matching) is decreased between 15% to 20%. Theoretical investigation and experimental results show that our algorithm has a very good performance; therefore this new algorithm is very suitable for fast edge-based stereo applications like stereo robot vision. © World Scientific Publishing Company.
IEICE Transactions on Information and Systems (17451361) 86(2)pp. 316-325
This paper presents an efficient Hybrid Learning Algorithm (HLA) for Radial Basis Function Neural Network (RBFNN). The HLA combines the gradient method and the linear least squared method for adjusting the RBF parameters and connection weights. The number of hidden neurons and their characteristics are determined using an unsupervised clustering procedure, and are used as input parameters to the learning algorithm. We demonstrate that the HLA, while providing faster convergence in training phase, is also less sensitive to training and testing patterns. The proposed HLA in conjunction with RBFNN is used as a classifier in a face recognition system to show the usefulness of the learning algorithm. The inputs to the RBFNN are the feature vectors obtained by combining shape information and Pseudo Zernike Moment (PZM). Simulation results on the Olivetti Research Laboratory (ORL) database and comparison with other algorithms indicate that the HLA yields excellent recognition rate with less hidden neurons in human face recognition.
The reduction of the search region in stereo correspondence can increase the performance of the matching process, in the context of execution time and accuracy. For edge-based stereo matching, we establish the relationship between the search space and parameters like relative displacement of the edges, the disparity under consideration, the image resolution, the CCD dimensions and the focal length of the stereo system. Then, we propose a novel matching strategy for the edge-based stereo. Afterward, we develop a fast algorithm for edge based-stereo with combination of the obtained matching strategy and the multiresolution technique using the Haar wavelet. Considering conventional multiresolution techniques, we show that the execution time of our algorithm is decreased more than 36%. Moreover, the matching rate and the accuracy are increased. Theoretical investigation and experimental results show that our algorithm has a very good performance, therefore this new algorithm is very suitable for fast edge-based stereo applications like stereo robot vision. © 2002 IEEE.
IEICE Transactions on Information and Systems (17451361) 85(11)pp. 1859-1871
Finding corresponding edges is considered being the most difficult part of edge-based stereo matching algorithms. Usually, correspondence for a feature point in the first image is obtained by searching in a predefined region of the second image, based on epipolar line and maximum disparity. Reduction of search region can increase performances of the matching process, in the context of execution time and accuracy. Traditionally, hierarchical multiresolution techniques, as the fastest methods are used to decrease the search space and therefore increase the processing speed. Considering maximum of directional derivative of disparity in real scenes, we formulated some relations between maximum search space in the second images with respect to relative displacement of connected edges (as the feature points), in successive scan lines of the first images. Then we proposed a new matching strategy to reduce the search space for edge-based stereo matching algorithms. Afterward, we developed some fast stereo matching algorithms based on the proposed matching strategy and the hierarchical multiresolution techniques. The proposed algorithms have two stages: Feature extraction and feature matching. We applied these new algorithms on some stereo images and compared their results with those of some hierarchical multiresolution ones. The execution times of our proposed methods are decreased between 30% to 55%, in the feature matching stage. Moreover, the execution time of the overall algorithms (including the feature extraction and the feature matching) is decreased between 15% to 40% in real scenes. Meanwhile in some cases, the accuracy is increased too. Theoretical investigation and experimental results show that our algorithms have a very good performance with real complex scenes, therefore these new algorithms are very suitable for fast edge-based stereo applications in real scenes like robotic applications.
This paper introduced an experimental evolution of the effectiveness of utilizing various moments as pattern features in human face technology. In this paper, we apply Pseudo Zernike Moments (PZM) for recognition human faces in two-dimensional images, and we compare their performance with other type of moments. The moments that we have used are Zernike Moments (ZM), Pseudo Zernike Moments (PZM) and Legendre Moments (LM). We have used shape information for human face localization, also we have used Radial Basis Function (RBF) neural network as classifier for this application. The performance of classification is dependent on the moment order as well as the type of moment invariant, but the classification error rate was below %10 in all cases. Simulation results on face database of Olivetti Research Laboratory (ORL) indicate that high order degree of Pseudo Zernike Moments contain very useful information about face recognition process, while low order degree contain information about face expression. The PZM of order of 6 to 8 with %1.3 error rate are very good features for human face recognition that we have proposed.
Usually, the stereo correspondence for a feature point in the first image is obtained by searching in a predefined region of the second image, based on the epipolar line and the maximum disparity. The reduction of the search region can increase the performance of the matching process, in the context of the execution time and the accuracy. For the edge-based stereo correspondence, we obtain the relationship between the maximum search space in the second image and the parameters like relative position of the edges, the disparity under consideration and the focal length. Considering the maximum of the disparity gradient in the real scene, we formulated the relation between the maximum search space in the second images with respect to the relative displacement of the continuous edges (as the feature points) in the successive scan lines of the first images. Then regarding to the pdf of the disparity gradient, we obtain maximum of the disparity gradient based on the parameters like disparity under consideration and focal length. Finally we developed some very fast stereo matching algorithms, based on the normalized cross correlation criteria (NCC) for different sizes of the matching block.
International Symposium on Image and Signal Processing and Analysis, ISPA (18455921) 2001pp. 164-169
Traditionally, finding the corresponding points has considered to be the most difficult part of stereo matching algorithms. Usually, the correspondence for a feature point in the first image is obtained by searching in a predefined region of the second image, based on the epipolar line and the maximum disparity. The reduction of the search region can increase the performance of the matching process, in the context of the execution time and the accuracy. We proposed a new matching strategy to reduce the search space for the edge-based stereo correspondence algorithms. Considering the maximum of the disparity gradient in the real scene, we formulated the relation between the maximum search space in the second images with respect to the relative displacement of the continuous edges (as the feature points) in the successive scan lines of the first images. Then we developed some very fast stereo matching algorithms, based on the nonhorizontal edges as feature points, and the normalized cross correlation criteria (NCC) with different sizes of the matching block (as the similarity measures). We applied these new algorithms on the Renault stereo image and compared the result with those of a traditional matching algorithm (20 pixels search regions and NCC with size of 15×15). The speed up of these new algorithms is between 2.8 to 13.8 and the percentage of errors is between 0.5 to 5.4.
In the gradient based learning algorithms, the momentum has usually an improving effect in convergence rate and decreasing the zigzagging phenomena but sometimes it causes the convergence rate to decrease. The Parallel tangent (Partan) gradient is used as deflecting method to improve the convergence. In this paper, we modify the gradient Partan algorithm for learning the neural networks by using two different learning rates, one for gradient search and the other for accelerating through parallel tangent, respectively. Moreover, the dynamic self adaptation of learning rate is used to improve the performance. In dynamic self adaptation, each learning rate is adapted locally to the cost function landscape and the previous learning rate. Finally we test the proposed algorithm, called accelerated Partan on various problems such as xor and encoders. We compare the results with those of the dynamic self adaptation of learning rate and momentum.
