Font Size: a A A

The Research Of Image Segmentation Based On Fuzzy Clustering

Posted on:2005-05-02Degree:DoctorType:Dissertation
Country:ChinaCandidate:A H ZhangFull Text:PDF
GTID:1118360152969123Subject:Computer system architecture
Abstract/Summary:PDF Full Text Request
Image segmentation is just to segment an image into different sub-images with different characters and get some interested objects. It is a key step from image process to image analysis, plays an important role in image engineering, and is applied in a lot of fields such as computer vision, pattern recognition, medical image and so on. Although many researchers have done a lot of work, presented thousands of approaches on image segmentation. Unfortunately there is no universal method which could be used everywhere. There is even not an objective standard for evaluating segmenting algorithm. The clustering-based method is very important and wide-used in image segmentation. The most common method of clustering analysis for image segmentation is the FCM (Fuzzy c-means Clustering), which doesn't need setting any threshold, or get people involved. It is very significant for the automatization of image segmentation. However, there are a lot of problems and difficulties while using FCM for image segmentation, such as getting the number of clusters, setting initial cluster centers and membership functions, non-global maximum, high computation, non-continuous regions and over segmentation, et al. After a deep research and analysis of the actuality and problems of image segmentation based on fuzzy clustering, an approach called as ESFCM (Edge-based Semi-Fuzzy C-means Clustering) is presented in this paper, aims at solving the problems and difficulties when using FCM for image segmentation. The ESFCM gets all info of the edge of image using simple and sensitive edge detector; Finds out all regions surrounded by the closed edge of image using region-growing technique. According to the rule "the nearest is the first", all pixels in the edge part are put into their respective nearest regions and a rough segmenting based on these regions is completed. Let the number of the rough segmented regions be the number of fuzzy clustering, the centers of each of the regions in rough segmentation be the initial centers of clusters. The initial membership matrix can be got according to the relation between the pixels and the rough regions. The methods above can solve the difficult problem to get the initial parameters for fuzzy clustering, avoid of the blindness of initialization of centers of clusters and membership matrix, decreases the possibility of converging local minimum. Based on the inherent character that the fuzzy part in an image should be close to the edge of image and the part without edge should be clear, in order to decrease the computation, we presented the idea and algorithm of semi-fuzzy clustering that uses fuzzy membership function with the edge part and crisp membership function with the part without edge. It has been proofed that the iteration of ESFCM could converge a minimum. Besides, the concept of distance from edge pixel to clusters as the second character is used in clustering. In this way, the continuity of each segmented region has been improved, and the segmented region is closer to the real object. We analyzed the difference between gray and color image segmentation, compared the complexity between ESFCM and FCM for segmenting, the quality between the ESFCM and other common methods of segmentation with many real images. From the result of a lot of experiment, the validity of ESFCM has been tested. For the segmentation of color image, we discussed the selection of color model, and worked out a Sobel edge detector, which is simple and based on YUV model. Based on the principle of "nearest in position and least in difference of characters", a solution of pro-processing was presented with the situation that there might be some non-connective, fragmentary, meaningless and over-segmented regions. Also, we gave out the techniques in detail for the implementing of all difficult parts of ESFCM. As an application of ESFCM segmenting method, we discussed how to use the ESFCM method in the Database-Image-Retrieval system based on interested region, and designed a prototype of the Database-Image-Retrieval...
Keywords/Search Tags:Image Segmentation, Fuzzy C-means Clustering, Edge-based Semi-fuzzy C-means Clustering, Edge Detection, Gray Image, Color Image, Image Retrieval
PDF Full Text Request
Related items