Image Segmentation Based On Level Set Method

Image segmentation is one of the most difficult and important technique in Digital Image Processing. In recent years, the algorithms based on level set evolution have been a hotspot in image segmentation study. In this paper, review of existing image segmentation methods is followed by a detailed introduction of level set theory. Then some improvements to image segmentation based on level set method are presented. Three proposed methods are introduced as follows:Firstly, an improved contour detection method based on level set and watershed transform are presented. The manner from coarse to fine is adopted in the method. The level set is primarily employed to segment the image and then watershed is employed to obtain accurate contours without over-segmentation.Secondly, aimed at natural texture images, multi-scaled sliding windows are applied to fit the textons so that their feature sets of probability density function are well obtained. In the end, a new evolution equation is derived and an improved image segmention method is gotten. Experimets show that our method has good results for natural texture images and SAR images.Thirdly, we present a segmentation algorithm for directional texture images. Features derived from Grey Level Co-occurrence Matrix and Stationary Wavelet Transform are used in our method. And experimental results show the validity of our proposed methods.