Genetic Algorithm For Image Segmentation And Pulse Noise Detection Methods Improve

Image segmentation is a hot research field of image processing. It would be difficult to have a segmentation technology appropriate to all the images because there are many types of images and each category have its own features. It is targeted at specific types of images corresponding research segmentation. The numbers of existing image segmentation methods are more than 100 including the method of thresholds with merits of simple algorithms and better segmentation quality which is widespread.Genetic algorithms (GA) as overall efficient parallel search methods are broadly applied in many subjects and areas. Genetic algorithms are suitable for large-scale search of an excellent space due to GA's robust, parallel and adaptive characters. Genetic algorithms are also gaining attention in the fields of computer vision applications. One example is that genetic algorithms are achieved good results in application of image segmentation. But the basic genetic algorithms is easily ripe due to impropriate fitness function designing and improper parameter setting.Image denoising is always a hot problem in image procession. There are many types of noises. Impulsive noise is a type of random nature and characters of non-continuous noise. Impulsive noise seriously affected by the presence of the image quality. Algorithms based on the noise detection have the merits of simple and fast on impulsive noise removal. But algorithms exists need to determine thresholds and constantly adjust the parameters, which making algorithms applied to the less convenience.This paper is targeted at the genetic algorithms and images of pulsating noise detection algorithms deficiencies, the two main research questions are:1) this article presents a genetic algorithm to improve and optimize images threshold segmentation results.. To maintain the diversity of species, to prevent premature convergence phenomenon, the paper use adaptive cross-rate and adaptive rate of variance whose values can auto change. The proposed algorithm successfully prevents premature convergence phenomenon. Experiments have shown that theimproved algorithm is faster than traditional methods in image thresholding segmentation.2) At the conclusion of the classic images denoising methods and new methods, based on noise detection algorithms shortcomings, this paper introduces one concept named change rate of gray level and applies it in noise detection. The new algorithm uses adaptive threshold to remove impulsive noise in one image. Proved by tests, the new method has good effects in image denoising. At last, two methods based on change rate of gray level are proposed to enhance images appearance.