Multilevel Residual Motion Compensation for High Efficiency Video Coding

摘要

The motion estimation, which is combined with the motion compensation in the decoding stage, is a fundamental method to achive high compression efficiency in video coding. The method achieves high compression efficiency by reducing temporal redundancy between adjacent pictures. Temporarily adjacent residual pictures also contain some redundancy even though the motion estimation technique removes some of the redundant information. Based on this characteristic, this paper proposes a multilevel residual MC (MRMC) algorithm, which repeats ME and MC on the residual signals in order to incorporate the possibility of further reducing the redundancy in the residual information. This paper mainly deals with the trade-off between the number of bits for residual signals and their corresponding motion vectors (MVs) based on the rate-distortion optimization. In addition, a fast method is presented for the practical use of the proposed MRMC method. The proposed method is implemented on the High Efficiency Video Coding software test model version 16.0. Experimental results show that an average Bjontegaard-delta rate savings of 9.16% can be achieved at a cost of increase of 17.0% in encoding time and 2.3% in decoding time.