Fast Mode Selection and Disparity Estimation for Multiview Video Coding

Multiview video coding is one of the key technologies in 3D video applications. However, considerable computational complexity is existed in multiview video encoding. In this paper, we propose a fast macroblock mode selection and disparity estimation algorithm. Firstly, the proposed algorithm terminates early the modes search process according to the correlation of modes in the corresponding locations of reference views. Compared with the full search method, the proposed strategy can avoid searching some unnecessary modes. Secondly, the disparity prediction is adjusted by the temporal correlation of disparity vectors. A small search range is used for disparity estimation. Finally, experimental results show that 67.42% ~ 90.06% encoding time can be saved by the proposed algorithm with an average of 0.05dB PSNR degradation compared with JMVM7.0.