The Study On The Coding Of Intra-Frame Synchronized Stereo Video

With the development of electronic technology, three dimensional video technology becomes more mature and has been widely applied in many different areas. The intra-frame synchronized stereo video has become one of the most important stereo video formats due to the same bandwidth needed for the transmission with current HD video and the compatibility with HDMI 1.4. A frame of intra-frame synchronized video contains two sub-frames, corresponding to left viewpoint and right viewpoint, respectively. The two sub-frames are captured at the same time with high correlation. The work in this thesis is on how to use this kind of correlation to perform more efficient video coding for intra-frame synchronized stereo video.In this thesis, we firstly introduce the progression of video coding technology with a series of stereo video formats included, where the intra-frame synchronized coding format is mainly introduced. Then we analyze the algorithm of disparity compensate prediction, which is the key technology for multi-view video coding, to lay the foundation of the proposed coding method for multiple reference frames based inter-view prediction.To propose the multiple reference frames coding method based on inter-view prediction, we investigate the main contents and related characteristics of H.264 and MVC. The proposed interpolates the coded and reconstructed left-viewpoint picture into the reference frame list as the inter-view referenced picture. Then the inter-view/inter prediction coding method is applied to the right-viewpoint picture based on the multiple reference frame coding method. Meanwhile, some efficient coding feature is also implemented in it. Experimental results show that the proposed method can compress the intra-frame synchronized video efficiently. After applying the multiple reference frame based inter-view prediction mode, the coding complexity of right-viewpoint video will be increased to some extent due to disparity estimation and intra-view compensation, The corresponding coding time will also become much longer To solve this problem, we improve the fast mode decision algorithm for multiview video coding based on dynamic threshold and propose a dynamic threshold algorithm based on separated viewpoint for the intra-frame synchronized video. The algorithm will set different threshold for left- and right- viewpoint video to make the mode decision more accurate for the reason that the two video streams are combined in the intra-frame synchronized video. It also decreases the complexity of coding efficiency and loss of PSNR. Combining the algorithm above and the multiple reference frames based inter-view prediction coding method, the intra-frame synchronized stereo video could be coded more efficiently.