A Study Of Energy Based Image Segmentation

Image segmentation is one of the most important steps leading to the analysis of processed image data—its main goal is to divide an image into parts that have a strong correlation with objects or areas of the real world contained in the image. In recent years the computer vision community has produced a number of useful algorithms for image segmentation. The form of energy function is the main issue currently for its clear description of the required problem and separation of the optimum skills.Energy-based segmentation methods can be distinguished by the type of energy function they use and by the optimization technique for minimizing it. The majority of standard algorithm can be divided into two large groups: (1) Optimization of a functional defined on a continuous contour or surface. (2) Optimization of a cost function defined on a discrete set of variables.In our paper we do the research on the Graph cut model in (1) and the Level set model in (2).First, we introduce a new model based on the Graph cut to perform the image segmentation and data classification. It uses the random forest to learn the information from the user's initialization and get the edge weights of the graph. With the maximum flow algorithm, we get our segmentation result finally.Second, a fast implementation method of Chan-Vese model is proposed, which does not require numerical solutions of PDEs (Partial Differential Equations). The advantages of traditional level set methods, such as automatic handling of topological changes, are also preserved.Finally, we propose a region-scalable active contour model with tensor field, which can be used for segmentation of texture image or non-texture image. Our energy function mainly consists of a region-scalable fitting energy term that draws upon image information in local regions at a controllable scale in order to segment images with intensity inhomogeneity, and a fast Chan-Vese model energy term based on the structure tensor for texture segmentation.