Research Of Image Segmentation Algorithm Based On Fuzzy C-Mean

As one of the basic problems in the field of image processing, image segmentation is not only a key technology among automatic target identification technologies but also the basis of target feature extraction, recognition and tracking. Image segmentation technology has been widely applied in the fields of image processing and analysis, such as military images, remote sensing images and medical images and so on.Fuzzy clustering analysis image segmentation algorithm is a kind of unsupervised clustering algorithm, it divide the image into c classes. Because of the use of fuzzy and uncertainty theory, the segmentation effect of fuzzy clustering analysis algorithm consistent with human vision system, it has became an important filed of image segmentation algorithm analysis. But fuzzy c means clustering algorithm (FCM) has some defects when applied to image segmentation, such as, the image does not make full use of spatial information, the number of clusters is difficult to determine, calculate load and other problems. For these shortcomings many researchers have proposed difference solutions. Based on the study of these methods, we propose modified FCM algorithms in this paper, the main works are as follows:1. In order to reduce calculate load of spatial fuzzy clustering algorithm when applied to image segmentation, we proposed a method ,which uses two- dimensional statistical information of original image and its smoothing image maps the image from pixel space to two-dimension histogram feature space, the method is also applied to kernel spatial fuzzy clustering algorithm. Experiments show that the algorithm segments the noisy images quickly on the premise that does not affect the effectiveness of segmentation, and it improves the operational efficiency effectively. 2. Fuzzy clustering for image segmentation should be a pre-determined to the number of clusters, and the method that initializes the center point randomly is easy to increase the iteration time and get into a local maximum. In this paper, an algorithm that combining mean-shift and fuzzy clustering is proposed. Firstly, the numbers and the center points of clustering are determined by mean-shift algorithm. Then, we segment the image with fuzzy clustering algorithm according to the information obtained by mean-shift. Experiments show that the method is a kind of fuzzy clustering algorithm which is capable of classifying automatically and is more stable for image segmentation.