Perceptual Fast Mode Decision Algorithm for AVS2 Intra Coding

Kaiyang Liu,Hao Xie,Guoqing Xiang,Rongbing Yang,Yuan Li,Wei Yan
DOI: https://doi.org/10.1109/ICSIP49896.2020.9339322
2020-01-01
Abstract:AVS2 is a kind of video coding standard independently developed by China. AVS2 achieves the same level of compression efficiency with the most popular international standard HEVC/H.265. However, the increased compression efficiency brings a significant increase in coding complexity. In order to reduce the coding complexity of intra coding, a number of fast mode decision algorithms have been studied. However, traditional algorithms mainly focus on maintaining the objective quality of the video without perceptual consideration. Therefore, a perceptual fast intra mode decision algorithm is proposed in this paper. Firstly, the Sobel operator is used to obtain the perceptual texture information of the current PU (Prediction Unit) with JND (Just noticeable distortion) model to determine the available candidate intra predicted modes. Furthermore, the predicted candidate modes are updated by spatial-temporal neighboring information and the main perceived texture direction. The experimental results show that the proposed algorithm reduce encoding time by 16.93% under all intra testing configuration on RD17.0, while the BD-Rate loss is 1.07% with SSIM metric.
What problem does this paper attempt to address?