Research On Automatic Image Segmentation Based On Superpixels And Graph Cuts

Image segmentation technique is an important part of pattern recognition, image analysis and computer vision etc, and its research has been attended by many scholars at home and abroad. Image segmentation uses the information contained in the image and conditions given by user to segment image into areas with different characteristics, and extracts objects area of image. In recent years, image segmentation technology becomes the focus in the image process. More and more academic theory has been used in the image segmentation and Graph Cuts, interactive segmentation algorithm based on Graph theory, is one of them. In order to further accurate object segmentation, star prior information is added into the Graph Cuts during segmentation process. However, interactive algorithm cannot cost too much in segmentation of sequence images. This paper designs an automatic image segmentation algorithm, by applying characteristic points obtained by superpixels algorithm and salient algorithm to star prior algorithm to complete the segmentation of the image.The first part is the research background and research significance of image segmentation algorithm, and introduces the technologies of image segmentation.The second part is star prior segmentation algorithms, including the single star and the star, and gives combination of shape prior and Graph cuts algorithm framework.The third part is automatic foreground points detection algorithm and extends star prior algorithm to automatic image segmentation algorithm.Last part is a lot of contrast experiments, it can be seen through the experiment, this algorithm has good performance.