This paper presents an image segmentation fuzzy model and algorithm based on multi-dictionary learning.In the proposed model the conformity within the segmented regions and the regularization of the boundary are considered by combining multi-dictionary learning and fuzzy method.On one hand

image patches are reconstructed by using the block means within the segmented regions and a structured dictionary with class labels.The class-specific reconstruction residual and the

l

2 regularization term measure the conformity within the segmented regions.This measurement can describe intensity information and texture pattern for different regions of images.On the other hand

the wavelet sparsity regularization is employed to preserve geometric shape of the seg

mented regions.Based on the alternating direction method of multipliers and dictionary learning method

we design a fast alternative iteration algorithm to solve the proposed model.In the proposed algorithm each step except wavelet shrinkage is a closed form.Hence it is easy to use.Numerical experiments are presented to demonstrate the efficient performance of the proposed algorithm.