Intra Prediction Fast Algorithm in AVS3 Based on Image Texture Characteristics

Yizhao Wang,Chaobo Zhang,Songlin Sun
DOI: https://doi.org/10.1109/iscit52804.2021.9590620
2021-01-01
Abstract:AVS3 Standard is China's latest generation of video codec standard that has been developed by China's AVS working group. Compared with the previous generation standard AVS2, AVS3 introduces a lot of new coding techniques, but also introduce a lot of coding complexity. To reduce the coding complexity of AVS3 video coding standard, this paper proposes three fast algorithms for intra coding. In the block partition stage, Kirsch operator is used to detect the image complexity in 64x64 CU, and different split depths are allocated according to this complexity; In the intra mode decision-making stage, this paper uses Kirsch operator to detect the image edge direction in each leaf CU node, so as to reduce the number of intra modes to traverse in the RMD process. At the same time, this paper uses this image edge direction to restrict the intra DT partition. Through these fast intra prediction algorithms, we can get a considerable intra coding speed improvement at the cost of less coding performance degradation.
What problem does this paper attempt to address?