Research Of Adaptive Differential Evolution Fuzzy Clustering Segmentation Algorithm

Image segmentation is the hot research topic in image processing, with the sharp increase of computer processing speed and the development of the real life needs, the color information of gray image is less, it can't meet the demand of people's lives, thus the color image segmentation is more and more aroused people's concern. But the threshold segmentation , edge segmentation, region growing segmentation, clustering segmentation method, on the edge of the image segmentation prone to discontinuous and scatter more problems. In view of this problem,this paper studies the adaptive differential evolutionary fuzzy clustering algorithm. This paper mainly includes the following three aspects.(1) In order to eliminate the impulse noise in color image, an adaptive median filter algorithm based on x-shaped window is studied, and the algorithm is compared with the adaptive median filter algorithm.(2) Because the fuzzy C mean (FCM) clustering algorithm divides the image, it is sensitive to the initial value of the clustering algorithm and is easy to fall into the local extremum. In order to overcome this problem, this paper studies the weighted FCM algorithm. Then, we use the global search optimization ability of the adaptive differential evolution algorithm to obtain the clustering center of the weighted FCM algorithm and update the membership matrix of the pixel. Finally, by optimizing the objective function, we obtain the membership degree of each pixel point to all the cluster centers, and determine the membership of the pixels to automatically classify the pixels.(3) In order to test the effectiveness of this method, four representative color image segmentation methods are selected to compare and analyze the segmentation results and performance indexes.Experimental results show that the adaptive median filter algorithm based on x-shaped window maintains the sharpness of the pixels, better protects the details of the image and the boundary information, and improves the noise reduction capability. At the same time, the improved adaptive differential-weighted weighted fuzzy C-means algorithm effectively suppresses the generation of local extrema and the background noise of the image, the effect of texture detail on the image segmentation effect, and overcomes the sensitive problem of the initial value selection, Segmentation of the integrity of the border, effectively enhance the segmentation effect,is a more efficient method.