An in Depth Comparison of Four Texture Segmentation Methods

Abstract

Texture segmentation is the process of partitioning an image into regions with different textures containing similar group of pixels. This paper presents a comparative study of four texture segmentation methods based on the following features: descriptors, heuristic function, fuzzy logic and Mask based features. Many types of textures are considered for analysis. The comparative results show that descriptor based approach is the most suitable for segmenting both natural and mosaic textures whereas heuristic function based approach is most suitable for random textures. Fuzzy features based approach is found to yield better segments for regular patterns while Mask feature based approach is the best for segmenting Natural images, but fails miserably on Mosaic textures. Fuzzy C-means classification is used for achieving texture segmentation.