A Fast Block Partitioning Algorithm Based on SVM for HEVC Intra Coding
Jun Yin, Yang Xu, Jucai Lin, Yao Chen, Ruidong Fang · 2018
Quad-tree-based coding unit (CU) block partitioning structure in High Efficiency Video Coding (HEVC) intra prediction causes a significant coding complexity increase. Hence, a fast block partitioning algorithm based on Support Vector Machine (SVM) is proposed in this paper. Firstly, some effective features are extracted from CUs in each depth as the input vector of SVM. Secondly, three offline trained SVM CU splitting models are loaded in each CU depth, which predict the class label of the current CU according to the extracted features. Moreover, the parameters of SVM models are resolved by grid search method. Finally, based on the predicted class label, the encoder will decide whether to split the current CU or not. Experimental results show that the proposed algorithm reduces the computational complexity of HM13.0 to 30.1% and 53.9% in encoding time with and without RDO(rate distortion optimization), while the loss in coding efficiency is negligible.