A Fast Level Set Segmentation Method Based on the Overall Information of Image
Min Li, Xiangmin Xu, Qian Min, Wang Zhuocai · 2010
Considering the characteristic that the nucleus and the nucleolus have no obvious boundary, this paper presents a fast level set method which is based on the overall information of an image to extract the nucleolus from the prostate nucleus. We pre-process the three-dimensional colorful prostate image at first, making the nucleolus (target) distinct to the surrounding environment (background) at the maximum extent. Secondly, we use the level set method to separate out nucleolus from the nucleus. The proposed method combines the edges information and regional information of the image, and it only considers the information of pixels which are around the zero-level set function in each iteration. Thus we not only have no need to re-initialize the level set function, but also have certain segmentation ability to the edge blurred images. Meanwhile, the method can improve the computation speed, too.