Researches On The Key Techniques In 3D Video Coding

With the rapid development of multimedia techniques, three dimensional (3D) video is becoming an important application in the everyday life. Compared with traditional two dimensional (2D) video, 3D video can provide users more realistic stereo experiences. Due to the advantages of 3D video, it will probably dominate the video market in the near future. In 3D video applications, bandwidth or storage consumption is still an important concern because of the huge amount raw video data. Therefore, efficient 3D video coding algorithm is highly desired for 3D video applications. The intrinsic characteristics of 3D video, as well as its application scenarios, bring some new problems and challenges, and also lead to the development of some new techniques of 3D video. Consequently, researches on 3D video are of both theoretical significance and practical value.In general, 3D video can be represented by multi-view texture videos and their associated depth maps. Based on such kind of 3D video data representation, this thesis investigates several key techniques in 3D video. The main contents and novelties of this thesis are as follows:1. This thesis proposes a new 3D video coding algorithm based on scalable video coding extension of H.264/AVC.In 3D scene, the color information and depth information are represented by the texture video and depth map, respectively. When the correlation between texture video and depth map is exploited, the compression efficiency may be improved compared with encoding them independently. A new encoding algorithm is proposed which employs Scalable Video Coding (SVC), the scalable extension of H.264/AVC, to compress the texture video and its associated depth map. Moreover, the proposed method complies with the SVC standard. Hence the generated bitstream can be decoded by the standard SVC decoder.2. This thesis investigates the impact of depth map coding quality on view synthesis.It is important to investigate the impact of depth coding quality on view synthesis for 3D video bit allocation. Depth map with too high bitrate will cause a bandwidth waste problem if it has little contribution to view synthesis. On the other hand, depth map with too low bitrate will make synthesized view's quality degrade if it is not sufficient for view synthesis. The thesis studies the correlation between depth map coding quality and view synthesis results. The experimental results show that the depth map quality has little impact on view synthesis, and depth map with low bitrate is able to provide excellent view synthesis results. Therefore, in 3D video application systems, depth map can be allocated with small amount of bits, which is sufficient for view synthesis.Experimental results indicate that the proposed 3D video coding algorithm can bring much more performance gain than traditional coding methods. Moreover, the investigation of the impact of depth map coding quality on view synthesis in this thesis can help to guide the bit allocation during depth map coding.