Active Contour Model And Its Application In The Fundus Image Segmentation

Segmentation technology of fundus image has long been a research focus and challenge in the field of image processing to assist the medical diagnosis. To some extent, the distribution of the specific structure in the fundus, such as optic cup and optic disk and the change of shape parameters can be indicators of the degree of pathological changes of certain diseases, such as glaucoma, macular degeneration, and diabetes. Such an indication can contribute to early diagnosis, which was the key to cure and control.This paper investigates the active contour segmentation model based on the region, and strives to seek breakthroughs in key technologies of computer-aided diagnosis to satisfy ophthalmic disease research needs. And segmentation of fundus images with high accuracy, repeatability and automation are finally realized. This research work is divided into four:1、We first review the traditional methods of image segmentation and sum up the current research in image segmentation technology. We then point out the advantages and disadvantages of these algorithms and outline the evaluation standard for medical image segmentation.2、To avoid the defects of the Snake model in the parametric active contour model, we introduce the geometric active contour model and analyze level set method theoretically.3、With the fact that early level set method is difficult to achieve good results, we study the Mumford-Shah model(M-S model), which is based on regional information model, the Chan-Vese model(C-V model), and the C-V multiphase active contour. And with this method, the multiphase segmentation of fundus images is reached, achieving the simultaneous segmentation between the optic disk and optic cup of ocular fundus images.4、We propose a new ellipse constrained multi-phase active contour segmentation model. Based on the different brightness of the gray images in the optic disc and optic cup, we develop a multiphase active contour model. The new algorithm can segment the optic disc and optic cup simultaneously and exactly. Through validation of the algorithm to different medical fundus images, the experimental results show that the method can segment the optic disc and optic cup accurately. Therefore, this algorithm proves to not be in need of pretreatment, with quantitative shape description information.