A New Multi-threshold Segmentation Method Based on MHFFCM

Image Multi-threshold Segmentation techniques are the important contents of image segmentation, one typical algorithm of which is Fuzzy C-Means (FCM) clustering segmentation algorithm. The conventional FCM clustering algorithm is based only on special information and ignores the spatial distribution of pixels in an image. Large numbers of improved methods are put forward to conquer this limitation, but all of them increased the computation cost greatly while the segmentation effects are not improved evidently. At the same time, the conventional FCM selects the initial clustering centers randomly, which greatly increases the iterative count. A new method based on fast FCM algorithm and multi-histogram (MHFFCM) is proposed in this paper, which utilizes the special and spatial information adequately by analyzing many kinds of characteristics among different intensity levels in an image. The importing of Multi- characteristic makes the selection of thresholds possible and easy. Besides, a selection method of initial clustering centers based on intensity histogram equalization is presented in this paper, which can decrease the iterative count and shorten the runtime. Experimental results indicate that this method can improve the segmentation effects obviously and decrease the computation cost greatly.