Abstract

Although FCM is helpful to handle the problem of uncertainty by avoiding setting threshold, when it is used to segment an image, it usually omits space information. It leads to be sensitive to noise. To settle it, the paper proposed a new algorithm to reach image segmentation, which is called dynamic granular fuzzy clustering algorithm. It firstly constructs a character vector of two dimensions, in which one is grey and the other is neighborhood grey means. Equal relationship is set up based on them. Secondly, attribute weight coefficient between vectors is given, which define the distance and the new clustering algorithm is produced. Then by it we achieve image segmentation in unit granular layer. Finally, by synthesis of multi-granular layers, we reach final image segmentation. In order to certify the new algorithm, it is applied to image segmentation tests. The results indicate that it is more suitable for actual need, which not only take space relationship into account, but also provide new thoughts in image process.

Paper Details