Level Set Medical Image Segmentation Based On Fuzzy Clustering

Medical image segmentation technology is one of the key content of medical image processing and analysis, the goal is to process the region of medical image area interested, those pathological changes of tissue which need to be segmented, make anatomical structures similar to the greatest extent, providing theoretical basis for further diagnosis and treatment for clinician. This paper analyzes the deformation model and fuzzy clustering which based on medical image segmentation method, the CV model?OSTU algorithm?double CV model and FCM clustering algorithm on the basis of the related improvement and the organic integration. Specific works are as follows:1. Variation principle and the process of the closed curve evolution are as the first analysis, then embed the curve evolution in the energy function to achieve target segmentation, and the active contour segmentation model based on the parameters of are introduced emphatic:geodesic active contour (GAC) model, the local binary fitting model (LBF) model, and analyze its advantages and disadvantages.2. For medical image noise, fuzzy boundaries between different soft tissue and lesions etc, this article selects clustering algorithm, combined the Li Chunming model (Li) and two phase level set model (CV) with improvements. Which chooses the appropriate filter to denoise for medical image, then, uses the fuzzy c-means algorithm (FCM) to obtain the prior model of image. Energy function of improved CV model, which is used to get the a priori second image segmentation afterwards. Through the experimental simulation, the model can deal with the phenomenon such as high noise better, weak boundary of medical images, and can avoid the re-initialization effectively, more sensitive to the edge. The segmentation accuracy improved, noise suppressed effectively, the number of iterations and time reduced significantly, certain application value.3. In view of the uneven gray weak boundary phenomenon often happens in the real world images, may be even more difficult for image segmentation, the local binary fitting (LBF) of traditional model segmentation method need complicated mathematical model is set up, localization properties have also led to the model is easier to fall into local extremum, and require a longer number iterative algorithm, in order to overcome the disadvantages of the above problems, put forward another method of image segmentation, firstly the background of the blood vessel image by the morphological operation, make the image of the target area is enhanced effect. After the bilateral filter is adopted to after dealing with the strengthening of blood vessel image filtering, with the method of variance between (Ostu) after filtering of the image segmentation, select the best threshold generation into the Canny algorithm for segmenting again, the experimental results show that this method does not need complex mathematical model is established and numerical analysis algorithm, significantly reduce the iteration times, the uneven gray and fuzzy boundaries of vascular images can be better segmentation.4. For in practice for segmentation of medical images in addition to the target and background region, there are two or more target area, traditional Chan-Vese model generally applies only to two phase image segmentation, not well multiphase image segmentation. This paper proposes a modified double based on fuzzy kernel clustering level set segmentation of medical images, by using KFCM clustering algorithm to reduce image noise and the sensitivity of the double level set model, to improve the double level set model, after the clustering image segmentation. This method has good ability to suppress image noise, make full use of the image edge information, without having to initialize level set function, reduce the number of iterations computation and algorithm, and can effectively achieve heterogeneous region segmentation.