Hard-decision quantization with adaptive reconstruction levels for High Efficiency Video Coding

In this paper, a quantization scheme based on hard-decision partition and adaptive reconstruction levels is proposed for High Efficiency Video Coding (HEVC), the video coding standard currently under development, which has demonstrated approximately 40% bit saving compared to H.264/AVC. For each video frame, the residual signal after motion estimation is transformed and then quantized by a hard-decision scalar quantizer. The reconstruction levels of the quantization outputs are adaptively computed based on the statistics of the residual signals of previously coded frames, and are selectively transmitted to the decoder. Compared with the rate-distortion optimized quantization (RDOQ) method in HEVC, which is attempting optimal among all encoder-side quantization techniques, the proposed scheme addresses the decoder-side reconstruction problem and achieves further rate reduction in the low-delay low-complexity setting while the encoding computational complexity is significantly reduced, and the decoder complexity remains almost the same.