Research And Implementation Of Medical Image Segmentation Algorithms Based On Fuzzy Clustering

Image segmentation refers to a technology that separates a given image into several areas that have their own properties and extracts the interested objective areas. Medical image segmentation is the previous process of other medical image processing and pattern recognition such as the characteristic quantification, the feature matching, the three-dimensional reconstruction etc.. It also gives a support for the clinical diagnosis and adjuvant therapy. Medical image has the partial volume effect and the uncertainty belongs to some tissue areas, which determines the fuzziness of medical image. The image segmentation algorithm based on fuzzy theory introduces the concept of fuzziness into it using the membership indicates the proportion of each pixel in every pure partial volume. This idea has been widely used in the segmentation of MRI brain image and the most typical algorithm is the Fuzzy C-Means Algorithm.Fuzzy C-means (FCM) clustering algorithm is a classical one in fuzzy cluster analysis algorithms, using iterative optimization of the objective function to obtain the fuzzy partition data sets, and is good at convergence. FCM algorithm can avoid setting the threshold, solve a number of branches of the partition problem, and suit images with uncertainty and ambiguity. As an unsupervised clustering algorithm FCM algorithm does not need manual intervention and is very significant for the automatic of image segmentation. Therefore, the application of FCM algorithm to image segmentation, being a hot subject in image processing, has a certain practical value.On the basis of analyzing the research status and development tendency of medical image segmentation both in the domestic and overseas, this paper has a deep discussion about the Fuzzy C-means clustering medial image segmentation algorithm, including the clustering class number c, the fuzzy weighed index m, the iteration cut-off error ε and the influence of initial clustering centers on the algorithm. Directing at the problems of low segmentation speed and unclear segmentation of noisy image, we proposed two modified algorithms in this paper. One is to introduce the kernel function into the FCM algorithm, and the other is to introduce the bound term of controlling the neighborhood effect. Finally, we proposed a new algorithm that incorporates a kernel-induced distance metric and a penalty term that controls the neighborhood effect to the objective function. The simulated experiment results show that the proposed algorithm has obvious advantage in iteration number and robust to noise than the standard fuzzy image segmentation algorithms.