Feature based color stereo matching algorithm using restricted search

Abstract

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.