Image Segmentation Algorithm Based On An Improved Active Contour Model

Image segmentation has become increasingly significant in many image analysis techniques and a key technology in the pre-image processing.We can have access to a large amount of useful information during the later process of iamges if the images can be effectively segmentalized. But if an error occurred during the image segmentation process, it will not only affected the later operation, but also may had disastrous consequences. The researchers have done a large number of research in terms of Image segmentation in recent years and among these,there is a lot of classic algorithms. However until now, there is no unified algorithm and standard to segmentalize the images effectively. But among the existing image segmentation methods, active contour model segmentation algorithm has slowly become an image segmentation mainstream.we first introduce some common image segmentation algorithm, including basic theory and application of these algorithms, analyze the advantages and disadvantages of the common algorithms, highlights the origin of the focus of this study Chan-Vese (hereinafter referred to as CV model), mathematical theory based on the limitations of the early active contour models. As the classic active contour models, CV model relative to the early active contour type has many advantages, it is possible to deal with the image of the topology, no longer dependent on the gradient of the edge of the image target for ordinary image, curve effective evolution shows good segmentation. But if there is a intensity inhomogeneity CV model often get poor results, Often in the actual operation, CV model is not good in dealing with complex images. In addition, How to set the initial outline of the CV model for intensity inhomogeneity always have lots of problems. And during the CV model calculations, re-initialize the level set function often takes too much time. In terms of stopping the curve evolution, the reference information of CV model is no longer applicable. Therefore, it is necessary to propose an improved active contour CV model.In this paper, we proposed a new model which based on a curve evolution, the level set method, local and global statistical information. The new model includes three parts:global energy function, local energy function and adjustments. The new model which introduces the local statistical information can effectively segmentalize intensity inhomogeneity image. In addition, Joinning the penalty term in the adjustment item can effectively avoid the level set function initializes and save time of the model evolution. A curve evolution discriminant basis proposed in this paper can make the contour curve evolution stop at the right boundary of the object. Experiments show that the new model can speed up the evolution model, the initial contour of the place have stronger robustness, can effectively intensity inhomogeneity image segmentation.