Image segmentation is a process of separating an image into several disjoint regions, whose characteristics such as intensity, color,texture etc., are similar. Image segmentation is a very difficult and problem specific. There is no universal method to solve this problem. In this disseration, we take three typic problems: general image segmentation, white matter lesion segmentation from volumetric MR images, and cell image segmentation as our research topics. Main contributions in this paper can be summarized as follows. We proposed a novel pixon-based adaptive scale method for image segmenta- tion. The key idea of our approach is that a pixon-based image model is combined with a Marker random field model under a Bayesian framework. In our method, we introduce a new pixon scheme that is more suitable for image segmentation than the "fuzzy" pixon scheme. The anisotropic diffusion equation is successfully used to form pixons in our new pixon scheme. White matter lesions are common pathological findings in MR tomograms of elderly subjects. These lesions are typically caused by small vessel diseases (e.g., due to hypertension, diabetes). We introduce an automatic algorithm for seg- mentation of white matter lesions from volumetric MR images. In the literature, there are methods based on multi-channel MR images, which obtain good results. But they assume that the different channel images have same resolution, which is often not available. Although our method is also based on T1 and T2 weighted MR images, we do not assume that they have the same resolution (Generally, the T2 volume has much less slices than the T1 volume). Our rnethod can be sun- marized as the following three steps: l) Register the T1 image volume and the T2 image volume to find the T1 slices corresponding to those in the T2 volume; 2) Based on the T1 and T2 image slices, lesions in these slices are segmented: 3) Use deformable models to segment lesion boundarics in those T1 slices, which do not have corresponding T2 slices. About cell image segmentation, we propose a novel approach by combin- ing kernel-based dynamic clustering and Ellipsoidal Cell Shape Model. A priori knowledge about cell shape is incorporated in our method. That is, an elliptical cell contour model is introduced to describe the, boundary of the cell. Our method consists of the following components: (1) obtain the gradient image; (2) use the gradient image to get, the image points, which possibly belong to each cell boundary; (3) adjust the parameters of the elliptical cell boundary model to match the cell contour
修改评论