Research On Fast Algorithm For Ultra-high-definition Video Coding

The current multimedia technology and signal processing technology are constantly updated and developed,and the user's requirements for the visual experience of video are also continuously improved.The requirements include high resolution,low latency,and immersive experience.Among them,the embodiment of high resolution is Ultra High Definition(UHD)video.The resolution of UHD video is able to reach 4K or even 8K.In order to encode UHD videos more efficiently,video encoding standards are also being updated.After HEVC,the Joint Expert Group on Video Coding has introduced Future Video Coding(FVC)and Versatile Video Coding(VVC).FVC and VVC greatly reduce the coding rate required for coding,but the new coding structure introduced therein greatly increases the coding complexity and limits the application of video in real-time scenarios.Therefore,this paper proposes three fast coding algorithms for different coding standards and their extended applications.(1)A random forest-based fast intra coding unit(CU)partition algorithm for FVC is proposed.First,the texture features of images are extracted.Then,the random forest models are established to predict the CU partitioning result.Finally,the unnecessary traversal of split modes is skipped,and time saving is achieved.Experimental results show that compared with the original platform's algorithm,the proposed fast algorithm for FVC is able to decrease the average encoding time by 44.12%,with negligible coding performance loss.And the BDBR only increases by 2.61%.The approach can also save more than 20% of encoding time relative to state-of-the-art methods,with BDBR slightly increasing.(2)A fast CU size decision algorithm for VVC intra prediction based on support vector machine(SVM)is proposed.The quadtree with nested multi-types tree structure of CU is optimized utilizing the SVM models by predicting the directions of splitting modes.Additionally,effective features are proposed to distinguish splitting directions.The proposed algorithm achieves a significant time saving by 51.01% and the BDBR increased slightly by 1.54%,which reduce the complexity of VVC while maintaining its encoding performance.(3)A fast VVC intra algorithm for panoramic video is proposed.The latitude information of the panoramic video is fully utilized.The algorithm includes two parts,fast CU splitting and fast intra prediction mode selection.In the first part,SVM models are used to predict the splitting result of CUs.In the second part,latitude information is used to simplify the selection of angle modes.The experimental results show that the algorithm saves encoding time by 54.82%,and BDBR increases by 1.66%.