DEEP LEARNING BAESD INTRA PREDICTION FILTER IN AVS3

AVS3 is the latest generation of video coding standard in China, which corresponds to the international VVC video coding standard. At present, intra prediction is still one of the most important parts for AVS3 and VVC. However, the current intra prediction is still based on the linear copy of adjacent pixels, which inevitably limits the performance of intra prediction. Based on deep learning, this paper optimizes current intra prediction modes, including ordinary angle based intra prediction (AIP) mode and intra block copy (IBC) mode, and achieves performance improvement up to 1.36% and 2.42% for the above two cases respectively.