An Interactive Image Segmentation Method Based On Local Search And Global Search

Image segmentation plays an important role in the field of the image processing. That affects the subsequent image analysis and image understanding directly. Since the natural image contains variety complex features and information. Like colors, textures and shape information, however, the automatic image segmentation method does not have a good segmentation results. The late research in this field, due to the semi-automatic image segmentation method can provide the prior knowledge for computing in the computer system. That provided an easy way for user to input interactive information, and assist computer to retrieve a better target region. That usually called Interactive Image Segmentation Method.In order to retrieve a better image segmentation result with complex background more accurately, the super pixels algorithm, the region merging algorithm based on local search, the graph cut algorithm based on global search, and interactive algorithm, all these image processing methods are combined in this paper. Lately, an interactive image segmentation method based on global search and local search is proposed. Not only the method reduces the image processing time but also retrieves a better segmentation results.In this paper, we are going to deal with the image firstly by marking the information of foreground and background. The users can use lines of different colors to divide the foreground and background area to assist computer to process the image. By completing of the markingwork, we intended to do some the pre-work for image segmentation through the popular image preprocessing algorithm-super pixels. The greatest strengths of the algorithm were making the super pixels as the processing unit instead of normal pixels. And that significantly reduce the complexity what makes the subsequent calculation quantity and the time complexity of the overall algorithm. Secondly, we are combining those adjacent areas based on similarity, used region merging algorithm based on local search with the maximum similarity mechanism.For the situation of merging regions that similar regions contains different relation of region in space, we introduced method graph cut algorithm based on global search, remedy the defects caused by local search algorithm. And divide super pixels again through the overall consideration of the segmented image. That also applied network flow theory for maximum flow-minimum cut theory to cut the graph cut algorithm,using criteria of minimizing the energy function. Therefore, the computer system can mark the foreground and background info accurately.The experimental data indicates that the new proposed image segmentation algorithm has faster and better segmentation results for the complex background target. Also applicable to the images with complex background, and the algorithm is better than a single image segmentation algorithm has better use value.