Researchs On Coding Technology Of Three Dimensional Televison

With the development in the areas of image processing, three dimensional (3D) displaytechnologies and network infrastructures, three dimensional TV (3DTV) is expected to becomeone of the most dominant markets. However, the data amount of3DTV is at least twice as that ofmonoscopic video, so it is necessary to compress the3D video efficiently for storage andtransmission. In this paper, we are focusing on improving compression efficiency for corlor videoand depth video in3DTV.(1) Fistly, in this paper, a fast inter predication algorithm is proposed for B-frame encoding inHBP based stereo video codec. The encoding information of the most similar mocroblock (SMB)of current macroblock in the right view is utilize to speed up the reference selection and earlyterminate the inter predication. Experimental results show that the proposed fast algorithm can save67.88%~75.38%encoding time in comparison with JMVM7.0, while the algorithm hardlyinfluences the rate-distortion performance. Then, a code optimization algorithm is proposed basedon virual viewpoint rendering. Experimental results show that the proposed algorithm can save4.6%~17.9%encoding rate in comparision with JMVM7.0, while the average of PSNR increase0.18dB.(2) Multi-view view plus depth (MVD) is an effective format for3D scene representation,where view synthesis based on depth-image-based rendering (DIBR) is employed to generatevirtual views. MVD signals must be effectively compressed before transmission. However, weaktemporal correlation of depth map sequence, caused by inaccuracy depth estimation and capturing,seriously decreases its compression efficiency. Therefore, depth map sequence should be processedbefore being encoded. In this paper, we proposed a novel preprocessing algorithm to refinetime-inconsistent depth map sequence. Experimental results show that the proposed algorithm canimprove time-consistency of depth maps and save22%~50%encoding bit-rate, while the proposedalgorithm improves the quality of rendered virtual views.(3) In practice, depth information is stored in8-bit grayscale format, whereas the disparityrange for a visually comfortable stereo pair is usually much less than256levels. Thus, severaldepth levels may correspond to the same integer (or sub-pixel) disparity value in the DIBR baseview synthesis such that some depth distortions may not result in geometry changes in thesynthesized view. Fistly, Based on the above analyses, this paper proposes a no-synthesis-errordepth preprocessing algorithm. Experimental results show that proposed algorithm can save the bitrate ranging from14.17%to74.55%, while the synthesized virtual view by the proposed algorithm has the same quality with the synthesized virtual view with the original depth map. Finally, aresidue signal preprocessing scheme in video coding scheme is proposed based onno-synthesis-error model. Experimental results show that proposed algorithm can save the bit rateranging from10%to25%for I frame and it can save the bit rate ranging from5%~10%for allframes.