Design And Implementation Of Applying The Image Segmentation Algorithm Based On Improved Genetic Algorithm

Image segmentation is the most central and important step from the image processing stage to the image analysis stage. Image segmentation is a kind of work which takes consideration on the gray scale, color and geometry and other characteristics of images. After these process, it will divide images into several areas, each area has its own characteristics, and then picked out the parts we need from these areas. Commonly, we mainly uses three methods,as follows for image segmentation. They are threshold segmentation method, edge method and region segmentation method, etc. Among above methods, the most commonly used method is threshold segmentation method, in this method, what threshold approached we often call it OTSU. There is a common problem in the image segmentation method that we can’t always finish image segmentation in high quality by using OTSU method, because of the complexity of different images.This paper proposes an improved OTSU method.In this method, we introduce a new measurement on cohesion with the variate of the average variance of background and target. In order to avoid unsatisfactory distance between the background and target and get better effect of image segmentation, we introduce two kinds of average variance concepts to get ideal cohesion and balance the average pixels in each class.Next, this paper proposes a new image segmentation method based on improved genetic algorithm which has optimized the solution. Especially the adaptive mutation operator selection, it was brought in considering about the characteristics of genetic algorithm, and actual operation efficiency of the algorithm. Experiments show that segmentation quality of gray image with noise has improved by using the new algorithm. And with the improved scheme, the computation time is greatly saved.At last, this paper proposes an image segmentation method which combines the improved genetic algorithm with the improved OTSU method. Simulation experiments show that the algorithm is able to maintain the diversity of the group. And speed up the rate of convergence at the same time. And threshold computation time is about 63% shorter than the two-dimensional OTSU image segmentation method, and shorter about 30% than the basic genetic algorithm. The algorithm improves the stability of the global convergence of the algorithm, and the threshold range is controlled within 3 pixels. This method can be used in initiatory image processing of head target recognition. Experiments show that the improved algorithm is more practical than traditional algorithm in analyzing the profile of head target, the effect of image segmentation and computation time.