Research On Key Technologies In Stereo Video Coding

Stereo video is a very important developing direction in video technologies in future, compared with single-view video, stereo video usually has two views and it's data amount is much greater than single-view's, so efficient compression is very important. There exit two methods to compress stereo video, block-based method and block-based method. Block-based method has the merits of simplicity, robustness and being easy to be implemented in hardware, but blocking artifacts may exist and the visual effect of reconstruction image may be bad in low bit rates. Object-based method has attracted great attention as a alternative to block-based method which can decrease coding errors and has better visual effects. Furthermore, it can describe a scene in a structural way and open new areas of application.In this thesis, the key technologies in the process of stereo video application are mainly studied. First, in chapter 2 block-based stereo video coding technologies are researched. Based on the analyzing of the constraints of disparity distribution and corrections among disparity vectors as well as corrections between motion vectors and disparity vectors, fast algorithms for disparity estimation are proposed. Experimental results show the proposed algorithms are efficient. This is one contribution of the thesis. During chapter 3 to chapter 5 object-based technologies are researched. In chapter3 object segmentation and tracking technology is mainly studied and one spatio-temporal based algorithm for object segmentation and tracking is proposed based on improved watershed algorithm and temporal information. The steps are as follows: First, motion area is extracted based on high-order statistics detection method and motion information of multiple frames. Second, the motion area and its surrounding areas are segmented with improved watershed algorithm. In the third step, the segment results are projected into motion area and the moving object is finally extracted after post-proposing. Experimental results show that for such videos as in conference environment with still background and simple objects, the proposed algorithm can achieve good results. This is another contribution of the thesis. In chapter 4 and chapter 5 object-based stereo and multi-view video technologies are mainly researched. First, algorithms for extracting moving objects in stereo and multi-view video are proposed. Then object-based disparity estimation technology is studied and two new algorithms for object-based disparity estimation are proposed, which can achieve precise and smooth disparity field. This is also a contribution of the thesis. For multi-view video, the intermediate image synthesis technology based on disparity interpolation principle is also studied which is essential to construct intermediate virtual image.At the same time, with fast development of digital technology and wide use of digital products, the protection for intellectual property rights of digital products is becoming more and more important. In Chapter2, we also studied digital watermarking technology which is developed in recent years is a new technology and is being used to digital products for copyright protection and product authentication. Present watermarking algorithms are mainly proposed for still images and audio signals, few for video sequences. Compared with still pictures, video watermark technology need high requirements. In the thesis, after analyzing of the features of video signals new watermarking algorithms are proposed in which watermarks are embedded into compressed video data-stream directly and the algorithms are very simple and efficient.