Study On C-V Active Contour Model

Image segmentation is a key process from image processing to image analysis, and is also a basic technique in computer vision. Its target is to isolate objects of the interest from an image and get the boundaries of the objects. There are great deals of researches on how to detect quickly, accurately and adaptively the objects in an image in the fields of medicine, military and industry. Recently, PDEs (partial differential equations) based segmentation techniques are gradually turned into research hotspot.One of the most popular geometric active contours models is C-V model (i.e., active contours without edges). The model does not utilize the image gradient and therefore have better performance than previous models. But it also there are many shortcomings, such as difficulty to deal with intensity inhomogeneity images, sensitivity to high noise, slow evolution.This dissertation focuses on C-V model. The main results are summarized as follows:①C-V Model with Edge Information. This model incorporates edge information into C-V model. It utilizes both the information of homogeneous regions and the edge information to stop the active contours on the object boundaries. The experimental results show that this model can overcome some disadvantages of C-V model, and obtain better results with respect to images that have the intensity inhomogeneity in objects or backgrounds, weak edges, and/or high noises while reducing significantly segmentation time. Besides, it has many advantages over other two improved C-V models (LBF and GACV).②Extended C-V model (EC-V). Intensity changes are crucial for accurate segmentation of many images; however, intensity changes are ignored in C-V active contours model. This extended version of C-V model has utilized intensity change information in images. The experimental results show: 1) EC-V model can be used to segment certain types of images to which C-V model is not applicable. 2) EC-V model is also able to segment the images that C-V model is applicable to, and is significantly fewer sensitive to noise than C-V model.