Fast QTBT Partition Algorithm for JVET Intra Coding Based on CNN

The latest Joint Video Exploration Team (JVET) employs quad-tree plus binary-tree (QTBT) block partitioning structure, which can improve coding performance significantly than HEVC with hugely increased encoding complexity. Aiming at alleviating the encoding complexity, we propose a novel fast coding unit (CU) depth decision method based on convolution neural networks (CNN) in this paper. Specially, the QTBT partition depth range is modeled as a 5-class classification problem, and design a single CNN classifier try to predict the depth range of CU directly, rather than to judge split or not at each depth level. The proposed fast scheme has the advantage that it can make the trade-off between the RD performance and time saving by setting different partition method; our “RD Maintaining” setting can achieve 43.69\(\%\) complexity reduction with only 0.77\(\%\) Bjontegaard Delta bitrate (BD-rate) increase. And our “Low Complexity” setting can achieve 62.96\(\%\) complexity reduction with 2.06\(\%\) BD-rate increase.