Based-on MPEG-4 Video Object Segment Arithmetic Investigation

Video coding standard MPEG-4 has a based-on content function which can be applied to some fields. To support this function and improve coding efficiency, one frame of video sequence should be decomposed to video object planes(VOP). It needs video sequence segment according to video object. The problem of base-on MPEG-4 video object segment is one of kernel problems of MPEG-4 and a very complicated problem. Video object segmentation technique is not a normative part of the standard. So I select the question for discussion of Base-on MPEG-4 video object segment investigation.Video sequence has too much data and Base-on MPEG-4 video object segment algorithm needs a good real-time character, So It is difficult that a normal computer completes real-time processing of video sequence. Because DSP kits have a powerful ability of real-time operation, these algorithm which had been proposed in this paper been proved base-on TMS320C6711 IDK(Imaging Developer's Kit) of Texas Instruments.The aspects of contents in this dissertation include as follows:1. The status in quo of video object segmentation and tracking algorithms are also presented, MPEG-4 standard is also introduced .2. Video signals are disturbed by all kinds of noises. They are harmful to our video segmentation. As we know, most of noises have higher frequency than signals. Therefore, low filtering can get rid of noises and protect video sequence edge characteristic. In this paper, we implement the low filtering on the platform IDK.3. To realize video object segment, we proposed one algorithm base on Bayes decision-making theory with least risk and video sequence edge information.4. To improve video sequence segment speed and video object continuity, It needs tracking video object. Base status in quo and characteristic of video object tracking, we proposed one video object tracking algorithm base on gray characteristic of video object region and contour characteristic of video object.