Fast depth intra coding based on spatial correlation and rate distortion cost in 3D-HEVC

摘要

3D-HEVC video coding standard adopts many sophisticated algorithms to improve the coding performance, which results in heavy calculation complexity. To reduce the computation burden, a complexity reduction scheme for depth map based on spatial correlation and the rate-distortion (RD) cost is proposed. Firstly, A maximum depth layer decision (MDLD) algorithm is proposed to predict the maximum depth layer (MDL) of each coding tree unit (CTU) accurately. Using the frequency distribution histogram theory in the statistics, the frequency distribution characteristics (FDC) of the RDcost J(0) of the CTU level in depth layer 0 is investigated under different MDL The initial threshold for different MDL is designed based on the FDC of the J(0) under allowable false rate (FR). An exponential relationship between the threshold and QP is discovered. Secondly, A fast depth intra mode decision (FDIMD) algorithm is proposed to skip unnecessary prediction modes. Considering the spatial correlation, the FDC of the RDcost J(ORM) of the optimal rough mode (ORM) in rough mode decision (RMD) is analyzed under different optimal prediction mode. Based on the FDC of the J(ORM) the mode-skipping threshold T H-st is designed to skip angular modes and depth modeling modes (DMMs). An exponential model is built to represent the relationship between the T H-st and QP. Finally, a fast wedgelet pattern decision method based on K-Means is developed to reduce the complexity of DMM1. Experimental results show that the proposed algorithm achieves an average reduction of 58.9% on intra coding time, with negligible degradation of coding performance. The proposed algorithm performs better in terms of complexity and performance than other approaches.