Swarm Intelligence Optimization Algorithm Of Multi-threshold Segmentation For Medical Image

Segmentation of medical image is the first,and toughest,challenge to medical science.The quality of segmentation will affect the operations of subsequent image registration and fusion.In the process of clinical diagnosis,surgery and radiotherapy,technology of medical image segmentation exhib its increasingly important clinical value.Otsu' method has been the most popular and basic segmentation technology because of the stable capability and simple implementation.With the increasing number of thresholds,the execution time of the algorithm will show an exponential tendency,so the application of multi-thresholding segmentation for medical image is still scarce.In order to solve the problem that multi-thresholding segmentation spends too much time finding the optimal solution in medical image segmentation.There are two new algorithms in this paper: Dynamic Combination of Genetic Algorithm and Ant Algorithm(DCG3A)merges with Otsu multi-thresholding merge.It uses genetic algorithm to generate preliminary partition results,and converts them to the initial information of ant colony.Then it uses ant algorithm to get the optimal values.In this process,it will check the conditions dynamically to avoid stopping genetic algorithm early or late,when combining the two algorithms.Besides,it increases the probability of variation and decreases the range when it checks the evolution rate.The experimental results showed that DC G3 A runs more quickly than the others,and it can always get the optimal values.Cuckoo Search algorithm is a new algorithm in recent years,and it is based on the habit of cuckoo to form a new heuristic algorithm.C uckoo Search algorithm in the search process introduce the lévy flight mode to search for space.It proves that the algorithm can find the optimal solution of global in the limited space,fastest and most efficiently.However,in the course of the experiment,we find that the initialization process of the search has some influence on the later implementation of the algorithm,so this paper proposes another algorithm :an improved cuckoo algorithm which is based on the initialization strategy.By introducing a unique initialization strategy to the cuckoo algorithm,the distribution range of the eggs is maximized and the local search ability of the cuckoo algorithm is improved.And set special termination conditions.The improved cuckoo algorithm is applied to Otsu multilevel threshold segmentation.Experimental results show that this algorithm can achieve fast and accurate convergence.