A non homomorphic block matching method is proposed based on Gauss mixture model and fast area belief propagation.The segmentation of multi scale adaptive fuzzy image is achieved

variable neighborhood search is taken based on the local contrast information changes

Bias method is used to estimate the fuzzy image edge energy characteristic

image region feature sub graph model is constructed

improved image segmentation algorithm of multi scale is obtained.The simulation results show that the algorithm can effectively improve the fuzzy image segmentation quality

shorten the time of segmentation

quality and stability is better than the traditional algorithm.It has better application prospect.