Moving Objects Based On Optical Flow And Level Set Methods And Multi-phase Segmentation

Motion segmentation is an important part of image segmentation. The goal of image segmentation is to separate image into different regions. And the goal of motion segmentation based on image segmentation is to separate the video or image sequences into different regions according to different standards. The main difference between image segmentation and motion segmentation is their object. As the high requirements of dynamic image, traditional image segmentation cannot satisfy the situation. So we need to do research work for motion segmentation. This paper mainly research motion segmentation of the video or image sequences, not only including two phases' motion segmentation but also multiphase motion segmentation model. Several aspects are introduced as follows: Firstly, this paper researches motion detection based on optical flow and improves date term and smooth term of optical flow model for enhancing robust optical flow field. Secondly, this paper researches two phases motion segmentation based on improved optical flow and Chan-Vese segmentation model, and adopts level set method to evolve curve. It will improve the two phases motion segmentation result's robust because of improved optical flow model. Thirdly, it extends two phases motion segmentation into multiphase motion segmentation, which will segment different objects including different motion information. Fourthly, all the models mentioned above are implemented. Many experiments denote that we get the satisfying results using the corresponding motion segmentation models. Lastly, future research directions are put forward according to the problems and limitation during research work.