A Method to Reduce the Intra-Frame Prediction Complexity of HEVC Based on D-CNN

Among a series of video coding standards jointly developed by ITU-T, VCEG, and MPEG, high-efficiency (HEVC) is one the most widely used today. Therefore, it still necessary to further reduce complexity HEVC. In HEVC standard, flexible partitioning procedure entitled “quad-tree partition” proposed significantly improve efficiency, which, however, leads high complexity. To intra-frame prediction, this paper proposes scheme based on densely connected convolution neural network (D-CNN) predict partition units (CUs). Firstly, block was designed efficiency CU fully extracting pixel features CTU. Then, efficient channel attention (ECA) adaptive kernel size were applied fast for first time capture information D-CNN channels. Finally, threshold optimization strategy formulated select best each depth balance computation performance RD. The experimental results show that method reduces encoding 60.14%, with negligible reduction in RD performance, which better than existing methods.