Early determination of intra mode and segment-wise DC coding for depth map based on hierarchical coding structure in 3D-HEVC

Abstract

This paper proposes a fast algorithm to reduce the high computational complexity of depth map coding in 3D-HEVC, which obstructs 3D video from the practical applications. Due to the special feature of depth maps, depth modelling modes (DMMs) were developed to keep sharp edges for depth map. Consequently, the 3D-HEVC encoder can improve the coding performance of synthesized views significantly. However, this is at the significant cost of increased complexity. Since 3D-HEVC reuses the well-known hierarchical coding structure, the mode information of the parent coding units (CU) can be reused to predict the most probable modes for their CUs. Besides, a method is also proposed for early determination of segment-wise DC coding (SDC) decision based on the hierarchical coding structure. Simulation results show that the proposed algorithm achieves a reduction of 33% encoding time while maintaining the quality of synthesized views in term of bit-rate values.

A reduced computational effort mode-level scheme for 3D-HEVC depth maps intra-frame prediction

“…The work [17], which was evaluated in 3D-HTM 8.1, achieved a timesaving of 27.9% at the cost of reducing BD-rate in 1.03%. The mode-level and a block-level combined scheme proposed in [18] saves 32.9% of execution time with the increase of 0.54% in BD-rate when evaluated in the 3D-HTM 13.0. The work [19] reduced 37.6% of execution time with an increase of 0.8% in BD-rate.…”

“…This section presents some published works focusing on mode-level reduced computational effort scheme/algorithms for intra-frame prediction of 3D-HEVC depth maps. Some works focus on accelerating RMD and MPM selection (e.g., [17]), some other works propose to diminish the encoding time on the intra-frame prediction of the TQ and SDC flows (e.g., [18] and [19]), and others target to reduce the DMM-1 wedgelet list evaluation (e.g., [17,20], and [21]). The work proposed by Zhang et al [17] is capable of simplifying RMD and MPM selection, and DMM-1 wedgelet list evaluation.…”

“…The work [18] performs mode and block level decisions. The modelevel decision always inserts into the RD-list the following modes: planar, DC, mode 10 (horizontal), mode 26 (vertical), and DMMs.…”

Overview of current development in depth map coding of 3D video and its future

“…In addition to these algorithms, some of our previous works [47][48][49] jointly reduce the number of CHIMs and DMM for residual coding with traditional RQT or SDC in steps 3 and 4. The candidate pool is pruned according to the costs from RMD [44,47], the reference pixels classification [48] or the hierarchical larger CUs [49].…”

Fast 3D-HEVC Depth Intra Coding Based on Boundary Continuity

“…Then, the number of DMMs was reduced according to the variance and the direction. By using the correlation between the current PU and parent PUs, Zhang et al [21] reduced the number of intra modes during RMD and determined whether to implement SDC. Hong et al [22] analyzed the rate-distortion cost (RD-cost) distributions of both rough mode decisions and DMMs with the one-sided Chebyshev's inequality to decide the adjustable threshold for early termination.…”