基于长边稀疏采样的AV1参考像素选择算法OA
AV1 Reference Pixel Selection Algorithm Based on Long-side Sparse Sampling
开放媒体联盟开发的第一代视频编码标准(Alliance for Open Media Video 1,AVI)作为新一代的开放式视频编码技术,以其高压缩效率和高质量视频输出等优势成为在线视频领域的主流编解码标准之一.旨在优化帧内预测编码技术以降低编码运行的计算复杂度,从而节约视频编码时间.为此,提出了长边采样(Long-side Sampling,LS)和长边稀疏采样(Long-side Sparse Sampling,LSS)的参考像素选择方法,根据块的划分方式规律地提取关键参考像素,以改进直流(Direct Current,DC)预测和亮度预测色度(Chroma from Luma,CfL)预测技术的性能.实验结果表明,LSS可以规避运算中的除法操作且减少大量加法计算量,在全I帧(All Intra,AI)和随机接入(Random Access,RA)配置下,相较原始算法平均编码时间减少11％和13％,同时节省0.03％和0.04％的Bjøntegaard Delta Bit Rate(BDBR).
As a new generation of open video coding technology,Alliance for Open Media Video 1(AV1)has become one of the mainstream coding standards in online video field because of its high compression efficiency and superior-quality video output.The intra-frame prediction coding technology is optimized to diminish the computational complexity of coding operations,thereby saving video coding time.To achieve this objective,two new reference pixel selection methods called Long-side Sampling(LS)and Long-side Sparse Sampling(LSS)are proposed,which systematically extract key reference pixels based on a block partitioning scheme to improve the performances of Direct Current(DC)and Chroma from Luma(CfL)prediction techniques.Experimental results show that LSS can avoid division operations and significantly cut off addition computations.Compared with the original algorithm,the proposed LSS algorithm attains 11％and 13％reduction in average coding time,0.03％and 0.04％Bjøntegaard Delta Bit Rate(BDBR)saving under All Intra(AI)and Random Access(RA)configurations,respectively.
黄婉溦;汪家华;刘昌沭;陈建
福州大学物理与信息工程学院,福建福州 350116
电子信息工程
视频编码开放媒体联盟开发的第一代视频编码标准帧内预测直流预测亮度预测色度
video codingAV1intra frame predictionDC predictionCfL
《无线电工程》 2024 (010)
2333-2338 / 6
国家自然科学基金(62001117);福州大学大学生创新创业训练计划项目(29218)National Natural Science Foundation of China(62001117);Innovation and Entrepreneurship Training Programme for College Students at Fuzhou University(29218)
评论