Finding
Paper
Abstract
The high efficiency video coding standard (HEVC) has come into view in recent years. In which some specific process and details were introduced to achieve higher coding efficiency, but lead to a very high computational complexity. In this paper, we focus on the need of lower complexity in computing recourse limited environment, and present a fast CU division algorithm and mode selection method for HEVC intra prediction. Experimental results show that the rate distortion cost (RD-Cost) values of no-split CU are clearly lower than that of splitting. So we formulate threshold equations of different depth level in the coding tree, by statistical analyzing of the RD-Cost values under different quantization parameter (QP) probabilities distribution. Division of code unit could be terminated earlier using threshold, which shall reduce the computing complexity. The experimental results show that, compared with HEVC testing model (HM), the improved algorithm can save an average 27.6% of encoding time with negligible loss of coding efficiency (only 0.49% bitrate increasing, and 0.013dB Peak Signal-to-Noise Ratio (Y-PSNR) loss). Meanwhile, by fully exploiting the correlation between the first rank prediction mode of the candidate mode set and the optimal prediction mode, some prediction modes can be skipped. It also leads to the decrease of computing complexity and the experimental results show that the proposed algorithm saves 42.1% of encoding time for the premise of ensuring the video quality.
Authors
Xiaoyan Xie, Xiaofei Xin, Hua Wang
Journal
2017 4th International Conference on Systems and Informatics (ICSAI)