The Study on Decision Tree Classification and Three-Step Search Algorithm Based Compression Technology for High Efficiency Video Coding

摘要

The new generation video coding algorithm HEVC reduces the bitrate by half compared to the previous standard H.264 with the same video quality, but this also brings the increasing of encoding time. In this paper, we mainly improve the intra frame prediction to solve this question: Combined with the decision tree method in machine learning, the reference pixel smoothing characteristics of the prediction unit of each partition mode are quickly determined, and some intra modes with the same prediction result are adaptively skipped, which greatly saves coding time. Experimental results show that the proposed decision tree and three-step search based intra prediction algorithm achieves better performance compared with the HM15.0 encoder. The BD-Rate is increased by 1.19%, the BD-PSNR is decreased by 0.058dB, and the encoding time is reduced by 37.87% on average.