Research On Medical Image Segmentation Algorithm Based On Fuzzy Multi-level Threshold

Medical image segmentation is an important branch of modern image segmentation technology.The quality of its results has a key impact on subsequent processing,such as organizational structure decomposition,disease diagnosis,and surgical implementation.Compared with natural image segmentation,the effect of medical image segmentation is often related to specific applications,imaging modalities and body parts,which is a more complicated problem.Many algorithms performing well in natural image segmentation have problems of under-segmentation or oversegmentation in medical image segmentation,and cannot be applied into the field of medical image segmentation directly.Therefore,it is necessary to study specific segmentation algorithms for specific application scenarios,imaging modalities and body parts.This thesis focuses on the multi-threshold segmentation algorithm and does the following two work.(1)Aiming at the influence of medical artifacts on the performance of segmentation algorithm,this thesis designs a fuzzy Kapur entropy based multi-threshold image segmentation algorithm(FKMTS).FKMTS introduces fuzzy Kapur entropy,which can effectively solve the problem of fuzzy boundary of organizational structure.In addition,FKMTS also uses an improved quantum particle swarm optimization algorithm to find the optimal segmentation threshold set.Simulation results show that FKMTS is superior to Otsu algorithm and maximum Kapur entropy algorithm in performance.(2)Traditional threshold segmentation algorithms are prone to noise and isolated blocks in the organizational structure.Aiming at this problem,this thesis designs a fuzzy Kapur entropy and neighborhood information based multi-threshold image segmentation algorithm(FKNMTS).Based on FKMTS,FKNMTS assigns the membership degree of different segmentation regions to pixels,and aggregates the membership degree of pixels and their neighboring pixels,thus improving the correlation of pixel membership in the neighborhood space.Simulation results show that FKNMTS has further improved the multi-scale structural similarity and peak signal-to-noise ratio than FKMTS.