Medical Image Segmentation Research Based On Deformable Model

Image Segmentation is one of the main research areas in the digital image processing theory system and plays an important role in the application of Computer Vision and Image Detection. However, there are still some difficult issues which need to be studied further, including weak boundary segmentation, multi-object detection, re-initialization and so on, which are the main topics of this paper.First, in order to increase the speed and precision of image segmentation, a novel gradient advanced segmentation model has been proposed based on Chan-Vese Level Set method, which not only eliminates the troublesome re-initialization process,but also attains the desired boundaries faster than others if there is more than one object in the detection area. Moreover, the gradient-advanced term can also increase the segmentation precision of weak boundaries.Second, aiming at the problem of multi-object image segmentation, a novel image segmentation method - Adaptive Multiphase Segmentation Model is proposed based on Chan-Vese model, which integrates the idea of multilevel segmentation and can calculate the number of Level Set curves according to the objects of images. This method eliminates the sensitivity of the initial curve in Multiphase Segmentation Model and decreases the segmentation step.At last, based on the result of the the fourth section, there are still some problems to be solved. Sometimes, it is difficult to use Var (Gray Average Variance) to control all the objects detection in images. So Aiming at the problem of multi-object image segmentation, a novel image segmentation method is proposed based on mutual information and Chan-Vese model, which integrates the idea of multilevel segmentation and introduces the concept of mutual information in information theory instead of gray average variance to decide whether the segmentation has been wholly completed. This improvement weaks the negative effect on segmentation results from noises and increase the segmentation precision of objects with weaks boundary.