Texture images segmentation by geometric active contour models based on color and gabor features

Abstract

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.