摘要
我国自主研究的音视频编码标准AVS3在AVS2的基础上引入了许多新技术,极大地提高了编码效率,同时提升了编码时间复杂度。为了推进AVS3编码标准的商用,必须研究降低编码时间复杂度,力求达到实时编码。为此,从图像频域特征角度提出一种基于图像DCT系数的帧内预测模式选择快速算法,在DCT系数反映出的图像纹理角度基础上结合统计算法预测得出当前PU块的纹理角度,从而减少帧内预测过程中的预测模式遍历。实验结果表明,本算法能带来平均39%的时间复杂度降低,且在高分辨率和纹理复杂的情况下综合性能更优,时间复杂度最高降低46%。
The audio and video coding standard AVS3 independently researched by our country has introduced many new technologies on the basis of AVS2,which greatly improves the coding efficiency and at the same time increases the encoding complexity.In order to promote the commercialization of the AVS3,it is necessary to study to reduce the encoding complexity and strive to achieve real-time encoding.This paper proposes a fast algorithm for intra prediction mode selection based on image DCT coefficients from the perspective of image frequency domain characteristics.This method combines the statistical algorithm to predict the texture angle of the current PU block based on the image texture angle reflected by the DCT coefficients,thereby Reduce the prediction mode traversal in the intra prediction process.Experimental results show that this algorithm can reduce the encoding complexity by an average of 39%,and the overall performance is better in the case of high resolution and complex texture,and the encoding complexity is reduced by up to 46%.
作者
康伏云
陈磊
许城钢
陶曦
刘子豪
KANG Fuyun;CHEN Lei;XU Chenggang;TAO Xi;LIU Zihao(School of Information and Communication Engineering,Beijing University of Posts and Telecommunications,Beijing 100876,China)
出处
《电视技术》
2021年第8期90-94,共5页
Video Engineering
关键词
帧内预测模式
DCT系数
纹理特征
intra prediction mode
DCT coefficient
texture feature
