RDNet: Rate–Distortion-Based Coding Unit Partition Network for Intra-Prediction

High efficiency video coding (HEVC) has been finalized as the most widely utilized standard, jointly developed by ITU-T, VCEG, and MPEG. In HEVC, quad-tree structure of unit partition is one substantial modules provides significant gains following huge time. this paper, a rate–distortion-based network (RDNet) proposed to make decisions based on statistical features. RDNet composed prediction sub-network target sub-network, where used predict CU modes intra-prediction designed optimize parameters evaluating rate–distortion cost, respectively. To balance accuracy loss, parameter-exchanging strategy applied control parameters’ sharing between two networks. Experimental results prove that our model can reduce encoding time HEVC 55.83~71.72% with an efficient BD-BR 2.876~3.347%, ablation study evaluates ability balancing trade-off inference speed.