Study On Image Segmentation Method Based On Fuzzy Clustering

Image segmentation is a key step from image processing to image analysis, is also a basic computer vision technology. The results of segmentation will affect the performance of high-level module in the image processing system, such as:analyzing, understanding and recognition etc. Therefore, the research of image segmentation has very important significance.Among lots of image segmentation methods, the segmentation method of fuzzy c means(Fuzzy c-means, FCM)clustering based on the pixel classification has been extensively researched, and has become a branch in the field of image processing. It is better than the traditional image processing methods. However, the standard fuzzy clustering segmentation is still some deficiencies.Firstly, it does not consider the inherent fuzzy information of the image. It is more sensitive to noise; Secondly, the image pixel space is large to make the longer time of the iteration computation. Therefore, it is an interesting research topic how to make effectively use of spatial information to improve the quality of segmentation, and without a substantial addition in computation.This paper offers the following improvements for shortcomings of image segmentation based on fuzzy c means clustering:(1) This paper presents a fuzzy clustering and image segmentation method based on high-dimensional feature vector and the pyramid structure. It not only takes advantage of inherent fuzzy information in image, but also makes use of the pixel gray information and neighborhood space-related information in image, gives the clustering segmentation method of the three-dimensional features and multi-scale analysis.Experimental results show that the method not only runs faster, but also has better segmentation effect in case of the interference of Gaussian noise.(2)The standard FCM algorithm is not only extremely time-consuming for clustering large data set, but also more sensitive to noise. For its shortcomings, this paper presents a weighted fuzzy clustering method based on three-dimensional histogram for image segmentation.It not only considers the gray-scale information of the pixels and the average and median value of the neighboring regions, but also gives different classified contribution to the different samples in the sample space. The experimental results show that the algorithm can meet effectively and timely to the image segmentation.