A Study Of Color Image Segmentation Method Based On Fuzzy Clustering

Image segmentation is the foundation of visual understanding, and also is one of the basic problems in computer vision, and has been applied in many fields. Because of the differences of image structure and content, it is still a difficult problem to achieve the rapid and general image segmentation. The main image segmentation research contents are:how to establish an effective segmentation model, reduce the complexity of segmentation algorithm, and improve the noise immunity and generality of segmentation algorithm. Clustering is an unsupervised classification method, which finishes classification by classifying samples of similar properties when we have no prior knowledge. Image segmentation method based on clustering analysis has small constraint to sample space, and have good generality. But it is still not perfect, mainly because of the large amount of computation, and the problem that it is easy to fall into local extremism. THSI article focuses on the study of fuzzy clustering algorithm for color image segmentation. It designed and achieved a FCM segmentation algorithm combined of edge detection techniques and improved rapid global K-means according to current research realization and the main problems.The main research worked in the following areas. Summarized and analyzed several current commonly used methods in color image segmentation, because the new algorithm committed to choose different methods to be combined according to the specific circumstances of the image. Classified and summarized a variety of color models commonly used in color image segmentation. Pointed out the meaning of every branch in these sorts color space, and the conversion to RGB color space, and their respective advantages and suitable occasions. The above work provided a basis for the choice of color space in color image segmentation. Pretreated the color image by median filtering to resistant noise in clustering segmentation. Deeply analyzed the principle of image segmentation method based on the fuzzy clustering, and corresponding studied several important clustering algorithms, which provide support for image segmentation. Then discussed several important parameters and the key problems to be solved in image segmentation algorithm based on fuzzy clustering, so it guide the improvement direction of the new algorithm. THSI paper presented a FCM algorithm based on edge detection and improved fast global K means. Extracted edge pixels through edge detection to compress the cluster samples in first round; used the improved fast global K-means clustering in edge pixels to obtain the initial clustering centers according to the problem that clustering algorithm is easy to fall into local extremism and results in less segmentation; used the initial clustering centers into the color images, in order to reduce the number of iterations, accelerate the convergence rate, reduce the amount of computation.Ran the improved algorithm in MATLAB, and experiments showed that the algorithm can effectively improve the operating speed of program without affecting the image segmentation result.