Multi-view video coding combined with adaptive prediction structure and fast mode selection

Abstract

The traditional multiview video coding (MVC) is based on hierarchical B picture (HBP) disparity/time hybrid prediction structure, making full use of the disparity and time correlation, largely improve the encoding efficiency, but also greatly increase the computing complexity. In this paper, an adaptive prediction structure and fast mode selection algorithm for MVC is proposed based on the analysis of the reference frame selection of HBP prediction structure coding and the statistical characteristics of macroblock (MB) mode selection. First, the disparity prediction of the current frame to be encoded is determined by the disparity, time correlation and time level of the multi-view video(MVV) sequence. Second, to execute fast mode selection by jointing rate distoRTion characteristic of mode selection and threshold value judgement. The experimental results show that the computing complexity of the proposed algorithm is reduced by 82.17% on average, compared with the default method set of JMVC (joint multi-view Video Coding Model) reference software-exhaustion mode selection-making, the Peak signal-to-noise ratio (PSNR) lost only 0.09 db and the total bit rate increased only 0.03%.