摘要: This paper describes an approach to perceptual segmentation of color image textures. A multiscale representation the texture image, generated by a multiband smoothing algorithm based on human psychophysical measurements appearance is used as input. Initial achieved applying clustering at coarsest level smoothing. The segmented clusters are then restructured in order isolate core clusters, i.e., patches which pixels definitely associated with same region. representing form 3D histograms for probabilistic assignment all other larger and categorise rest image. process setting up reassignment propagated through finer levels until full highest resolution.