| 摘要: |
| 为提升音视频编解码第三代标准(Third-generation Audio Video Coding Standard,AVS3)的编码效率,提出了一种使用帧级纹理信息的编码单元(Coding Unit,CU)划分和帧内预测联合快速算法,计算过程中,每帧纹理信息只计算一次,将其复用于编码的两个阶段,从而提高编码效率。首先,以帧为单位在图形处理器(Graphics Processing Unit,GPU)端计算图像的纹理信息,过程中利用积分图降低其计算复杂度;其次,根据CU的纹理信息及二叉树划分的编码代价,指导后续CU划分;最后,通过分层的帧内预测模式搜索,结合纹理信息,逐层确定最优预测模式。实验结果显示,与AVS3官方参考软件HPM15.0相比,在BD-Rate增加0.46%的情况下,所提算法平均减少了36%的编码时间,有效降低了AVS3的计算复杂度。 |
| 关键词: 视频编解码 编码单元划分 帧内预测 边缘检测 |
| DOI:10.20079/j.issn.1001-893x.250111002 |
|
| 基金项目:智能警务四川省重点实验室开放课题(ZNJW2023KFZD003,ZNJW2024KFMS003) |
|
| An AVS3 Partitioning and Intra-prediction Algorithm for Frame-level Texture Reuse |
| ZHANG Quana,WANG Shuna,LIU Yangyi,DUAN Changb,PENG Boa |
| (1a.School of Computer Science and Software Engineering;1b.School of Electrical Engineering and Information,Southwest Petroleum University,Chengdu 610500,China;2.Intelligent Policing Key Laboratory of Sichuan Province,Sichuan Police College,Luzhou 646000,China) |
| Abstract: |
| To improve the encoding efficiency of the third-generation audio video coding standard(AVS3),a joint fast algorithm for coding unit(CU) partitioning and intra-prediction based on frame-level texture information is proposed.The texture information of a frame needs to be calculated only once and is applied to two stages of the encoding process.First,the texture information of the image is calculated on the Graphics Processing Unit(GPU) side on a frame-by-frame basis,and the complexity of obtaining the texture information is reduced by using an integral image.Then,the subsequent CU partitioning is guided based on the texture information of the CU and the coding cost of the binary tree partitioning.Finally,through a layered search of intra-prediction modes combined with texture information,the optimal prediction mode is determined step by step.Experimental results show that compared with the official AVS3 reference software HPM15.0,the proposed algorithm reduces the encoding time by an average of 36 |
| Key words: video coding and decoding block partition prediction intra-prediction edge detection |