Moving Foreground Segmentation Algorithm In Video Sequences

With the development of multimedia technology, the second generation video coding standard has been the main aspect of research and applications. Compared with the pixel-based first generation video coding standard , the second generation video codingstandard is content-based. In this transformation,segmentation of moving foregrounds that people have intersting from video sequences is very important for second generation coding. In fact, segmentation is the basic prerequisite for content-based video applications, which contain video searches, coding, multimedia exchange, computer vision, etc. The effects of foregrounds segmentation will affect subsequent applications directly.About the foregrounds segmentation in viedo sequences, many works have been done . We also have a lot of algorithms but they all have their disadvantages. At the present time, there is no current algorithm which can segment foregrounds from the background efficiently, so foregrounds segmentation in viedo sequences has became the hot research area of image process.This paper introduces the basic theory of digital image process and the fundamental knowledge of iamge segmentation firstly, and discusses the existing algorithms of foregrounds segmentation. After analyzing the advantages and disadvantages of the algorithms, in order to improve the segmentation effects and the operation time ,an effective algorithm of foreground segmentation in video sequences is proposed.First, based on the character that the frames of the viedo sequences are different ,getting rids of making changing region by direct frame difference or accumulation of frame difference. We obtained the changing region by the intersection clustering of two kinds of accumulation of the consecutive and discontinuous frame-difference. So the algorithm can ont only get over the incompletely of the edge gotten by the frame diffreence but also can resolve the shadow and the disconvergence phenomenon gotten by the accumulation of frame difference. In this way, the algorithm can make the changing region converge to foreground's contour and get a better segmentation effect. The using efficiency of the frames can be improved greatly as well.Secondly, we get threshold by the LHS (least half samples) method,make threshold and morphological operations of the changing region, then we can get a comlpete foreground contour.At last, after binarization and twice scanning the foreground contour, we can get extracted moving foreground by filling moving region with original sequence.Experiment results show that the algorithm can successfully segments the moving foreground from the video sequences with small computational complexity, less time, low complexity. This algorithm which is a better moving foreground segmentation method is quite suitable for the viedo sequences in video conference, news broadcasting, people interview and so on.