Image Segmentation And Phase Unwrapping Based On Differential Immune Evolutionary Algorithm

Differential Immune Clone Algorithm as a kind of intelligent optimizationalgorithm, is used to find the global optimal results. The algorithm adopts twopopulation evolution at the same time and maintain the population diversity. Throughclonal reproduction, clonal selection, crossover and mutation operators makesalgorithm to the optimal solution direction continuously evolved, high parallelism ofthis algorithm can obtain the global optimal solution with fast convergence speed andlarger probability.This paper mainly research the Differential Immune Clone Clustering Algorithm.By using alternative or adding other algorithm to improve Differential Immune CloneClustering Algorithm and get more optimal algorithms. These algorithms proposed inthis paper have good performance in image segmentation problem and phaseunwrapping problem. The main work in this paper summarized as follows:(1) Immune Memetic Clustering Algorithm is proposed, this algorithm is based onDifference Immune Clone Clustering Algorithm. Differential Evolution Algorithmreplaced with Memetic Algorithm. The thought of two population evolutionsimultaneously adopted in the proposed algorithm. A population evolves with MemeticAlgorithm, and the other evolves with Immune Clonal Selection Algorithm. In nineimages segmentation experiments, the result shows this algorithm of two populationsevolution at the same time is better than single species in traditional algorithms.Compare with Differential Immune Clone Clustering Algorithm, the results shows thatthe algorithm has better performance in the global search, local search andconvergence speed.(2) A Rough-set and Differential Immune Fuzzy Clustering Algorithm is proposed.This algorithm is based on Differential Immune Clone Clustering Algorithm, andchange hard clustering into fuzzy clustering, because it can obtain more abundantclustering information. The advantage of rough set is process uncertain data. So thealgorithm with rough set fuzzy clustering thought is more conducive to solve theuncertainty problem. In experiments, nine images segmentation and other fouralgorithms compared experiments are used to validate the algorithm in the clusteringperformance stability, the advantages of the experimental results show that thealgorithm has higher segmentation accuracy and more accurate segmentation results. (3) Immune Clone Hybrid Algorithm is proposed. This algorithm is based onDifferential Immune Clone Clustering Algorithm, Differential Evolution Algorithmreplaced with Hybrid Genetic Algorithm. Using two populations evolutionsimultaneously to increase population diversity. The two populations were appliedImmune Clonal Selection Algorithm and Hybrid Genetic Algorithm for evolution. Thisalgorithm slove the problem of Hybrid Genetic Algorithm, which is early and poorstability can make the algorithm sometimes can not be obtain the satisfied result insolving branch cut phase unwrapping problem. Through the experiments in simulationand real phase maps, show that the proposed algorithm has more accurate phaseunwrapping solution and faster convergence speed compared with Hybrid GeneticAlgorithm method.The paper is supported by the National High Technology Research andDevelopment Program (863Program) of China(No.2009AA12Z210), the NationalScience Basic Research Plan in Shaanxi of China (2010JQ8023), the FundamentalFunds for the Cental Universities (K50510020011).