Moving Object Segmentation Algorithm For Video Sequences

The video segmentation plays an important role in the application of multimedia, and has a widely future in video coding, video browsing, multi-media interacting and computer vision. Video segmentation is the base of the content-based visual applications. The results of object segmentation will affect subsequent applications directly. At the present time, there is no current method, which can segment object models from the background efficiently, though a great deal of research work has been done for video coding. Most algorithms aim at particular image sequences. The video segmentation has been widely applied in many fields, especially in low bite-rate ratio multimedia fields, but there is a little algorithm about that, so it is more and more becoming the hot point in the video research field.This paper discusses the basic theory of digital image segmentation, and analyzes the exist method for the segmentation of moving objects in video sequences. Aiming at low bit rate video sequences, an effective moving object segmentation algorithm is proposed, and plenty of simulations have been carried out by this segmentation method.The basic idea of the method is: in the low bit ratio video sequence, video objects move slowly, there is a little change in the background and the inside of the video objects. However, the edge between video objects and background, to some extent, will have obvious changes among frames. But these changes are not reflected completely on the edge. So according to the statistical characteristics, if we use enough consecutive frames, the entire edge of moving object will be included on the edge of the accumulating frame difference. So in this method, Firstly, using accumulating frame difference, we obtain the exact moving object localization which makes up the incompletely of the edge gotten by the frame difference. Then we can obtain the frame difference mask by the threshold and mathematical morphology procedure. For sequences with a comparatively complex background, we can get the frame difference mask by combining temporal and spatial segmentation. "Canny" operator is used for spatial segmentation to find edges of images. Finally, the moving object mask is...