Fast Adaptive CU Partition Decision Algorithm for VVC Intra Coding

The latest video standard - Versatile Video Coding Standard (VVC/H.266) has been standardized and officially entered into force. Compared with the High Efficiency Video Coding (HEVC/H.265), owing to the introduction of the Quad-tree with Nested Multi-type Tree (QTMT) division mode, the encoder can choose a more detailed division type when dividing the Coding unit (CU), thereby improving the coding performance. However, when the CU selects the division type, it needs to traverse all possible division types and compute the Rate-distortion (RD) cost, which greatly enhances the coding complexity. Therefore, this paper designs a Joint Random Forest Classifier (JRFC) to make decisions on CU partition types, and proposes fast adaptive CU partition decision algorithm for VVC intra coding combined with our previous work. The algorithm has the capability to make partition decisions for CUs of diverse sizes (smaller than
$32\times 32$
) and completely bypass the intricate process of Rate-distortion Optimization (RDO), resulting in a significant reduction in encoding time. The experimental results demonstrate that, in comparison to the basic approach utilized in VTM10.0, the algorithm proposed in this paper reduces the encoding time by 57.27% on average, with only a marginal increase of 1.53% in Bjøntegaard delta bit rate (BDBR).