Multi-view video coding based on sequence correlation

Traditional coding method used the same prediction structure to encode the multi-view sequences whose correlation are distinctly different. This method eliminated the temporal and spatial redundancy by using multi-references blindly. Although the traditional prediction structures have quite good coding efficiency, it results in the problem of high complexity. So by analyzing the characteristic of the multiview video data itself, we designed a correlation testing structure and depending on the correlation intensity on both the temporal and spatial direction, a method of multiple prediction structures coding jointly is proposed. Experimental results show that the proposed method can reduce the coding complexity while maintaining high coding efficiency, and to the sequences whose spatial correlation is quite strong, the method given can not only raise the PSNR, but reduce the bitrate as well.