Research On Image Segmentation Based On Active Contours

Image segmentation has become an important image analysis technology. In the research and application of image, people often only interested in certain parts of the image, these sections often refer to objects or prospects.Objects and prospects correspond to the region with a particular and unique nature. In order to identify and analyze objects, we need to extract them from the image, on this basis, we could make further use of the image. As the basis of image analysis and understanding, image segmentation is one of the most basic but also difficult problems in the field of computer recognition. The existing image segmentation algorithms are special; these methods are aimed at specific images. Because of the diversity and complexity of image, there is no uniform segmentation implementation.Through the analysis of the domestic and foreign image segmentation research situation and development trend, this paper arrangements and summarizes the general theory and method of image segmentation, and make further research on segmentation algorithm based on actice contours. Our key research is active contours model which is applicable to heterogeneous images.Chan-Vese (CV) model is the most classical active contours model. C-V model bases on the theroy of boundary evolution and the method of level set, which can detect the object in the image by using an active contour. Active contour stop evoluting at the expected boundary. The force of evolution is decided by the global internal and external gray-scale mean of contours. It is a region-based algorithm. The outstanding feature of this model is that the object's boundaries to be detected are not defined by gradient. This feature reduces sensitivity of model to noise in image, but C-V model always generate wrong segmentation with heterogeneous image. In order to improve inaccurate segmentation when we use C-V model to segmente heterogeneous images, this paper propose a local property function to limit reference range of statictis information which is used by contour evolution, and import this function into active contours model to build a new model which is forced by local statistics information. When segmenting heterogeneous image, experiments show that according to the reference information of points which are on the active contour, the active contour can stops evoluting on the expected boundary and obtains the ideal segmentation results.