Research Of Medical Image Segmentation Algorithm Based On Fuzzy Clustering Theory

An important branch of digital image segmentation is medical image segmentation, which is an important early step of clinical medical diagnosis. Good medical image segmentation image can provide great convenience for doctor’s judgment about patients’ diseases. Therefore, research of medical image segmentation algorithm is currently a hotspot.This thesis mainly analyses the common medical image segmentation algorithms and theirs basic principle and segmentation results. Because of the fuzziness that medical image owns, we put forward fuzzy clustering theory to segment the medical image. In this paper, firstly, use ordinary c-means clustering algorithm (FCM) for medical image segmentation. Then, in view of the problem that existing FCM algorithm is sensitive to initial point, and easy to get into the local minimum value state, so a kind of fuzzy core clustering algorithm is introduced. By using nonlinear mapping, this algorithm optimizes the characteristics of the samples. By using mercer core, the samples of input space are mapped to high dimension feature space, and then cluster inside the feature space. So, compared with FCM, KFCM can extract features of samples better, so as to achieve better segmentation effect.In addition, based on the KFCM algorithm, this paper puts forward a dynamic-weighting fuzzy core-clustering algorithm. It adds a dynamic weighting parameter and can optimize clustering process, makes clustering effect of KFCM algorithm better, then achieves medical image segmentation better. Considering that compared to other segmentation algorithms, watershed segmentation algorithm is pretty accurate for segmenting each part of image. But, as the existed shortcomings of over-segmenting, and kernal clustering algorithm can merger over-segmenting parts resulted by watershed segmentation, so this paper puts forward a new medical image segmentation algorithm based on watershed algorithm and dynamic-weighting fuzzy kernal clustering algorithm, experimental results show that this algorithm can obtain better segmentation images.