Research Of Differential Evolution On Image Processing

Biological evolution generally followed Darwin's criterion: "natural selection, survival of the fittest", that adapt to the natural environment by the choice, crossover and mutation between individuals. Evolutionary algorithm is an optimization method that imitates biosphere evolutionary process. It doesn't dependent on the specific characteristics of problems. The advantage of Evolutionary algorithm is common, simple and parallel processing. Thus evolutionary algorithms are considered the key technology which has a significant impact on the 21st century computer technologies.Differential Algorithm is proposed lately. It becomes a hot topic in evolutionary algorithms for its strong global convergence, robustness and stability. Differential evolution algorithm reduce operational complexity by that it retains the global search strategy that based on population, uses real number coding, a simple mutation operation based on difference and one to one competition strategies. Because differential evolution algorithm is an efficient, simple parallel algorithm, so its oretical and applied research has important academic significance.Based on the theory research of differential evolution algorithm, this is given for different improved algorithms for different problem. Compared with OTSU method, using differential evolution algorithm can save a lot of time for solving image segmentation problems. In the segmentation problem of image with noise, the paper improved differential evolution algorithm by secondary search. That improves search capabilities in the later evolution stage the visual effect.Image restoration problem is one of the important issues in the Image processing. The main difficulty of Image Restoration is a large of information and too slow speed. Therefore, this paper uses fast convergence and stability of differential evolution algorithm in image restoration. In the process of image restoration, Algorithm achieves good results which carry on crossover and mutation operation in a randomly selected window based on the image characteristics.Finally, this article uses differential evolution Algorithm to solve image matching problem, and compares with gray correlation matching, SSDA. Experiments show that this algorithm saves considerable time, and matching accuracy is 90%. This paper puts chaos optimization methods and Differential Evolution together to settle image matching. Experimental results show that news method has faster convergence speed, but its accuracy decreases slightly.