Image Segmentation Algorithm Based On Active Contour Model

Image segmentation is one of the most critical techniques of the image processing and has great significance to image understanding, image analysis, computer vision and so on. Although there are a wide variety of theory, methods and algorithm, which have their own strengths and weaknesses, none of them can be successfully applied to all kinds of images. In this paper, we take attention to the active contour model applied to image segmentation.Active contour model is an effective means of image segmentation and received wide attention in recent ten years. It is the successful application of PDE in the field of image processing. The main idea of the active contour model is an energy minimizing active curve guided by external constraint forces and influenced by image forces that pull it toward features such as lines or edges. It can be divided into two categories, namely:the edge-based and region-based. The classic active contour models include Snake model, GAC model and C-V model. The contents of this paper are:(1) Introducing the development of active contour model, and the basic mathematical theory. Then we focus on the several classic models and some famous improved models. Some experiments and analysis are been done to show their results.(2) We research into the GAC model for segmenting textured images in depth. First, we introduce a method based Gabor texture space to segment textured images. It needs convolve the Gabor functions with the original textured image to obtain different channels. Some of these channels will be the input of the multi-channel GAC algorithm. And then we propose an improved multi-channel GAC model used to segmenting color-texture images. The new method improves the edge function of the original multi-channel GAC model, which has both the information of color edges and texture edges. Textural homogeneity can be characterized by J-Image which is the key of JSEG method. Experiment shows that the improved approach produces good results for color-texture images.