Human Understandable Features for Segmentation of Solid Texture
Résumé
The purpose of this paper is to present new texture descriptors dedicated to segmentation of solid textures. The proposed texture attributes are inspired by the human description of texture and allows a general description of texture. Moreover it is more convenient for a user to understand features signification particularly in a man-aided application. In comparison with psychological measurements for human subjects, our characteristics gave good correspondences in rank correlation of 12 different solid textures. Using these texture features, segmentation results obtained with the classical K-means method on solid textures and real three-dimensional ultrasound images of the skin are presented and discussed.