摘要： 提出基于边界纹理一致性的视频帧内模板匹配预测算法，根据图像内容纹理的一致性，在绝对差和(或均方差)的基础上，引入边界纹理一致性约束，形成新的模板匹配准则，以增强模板匹配预测与当前块周围像素之间的相关性，提高模板匹配预测的准确性。实验结果证明，该算法相对于帧内模板匹配预测算法，编码效率平均提升0.11 dB，最高能提升0.17 dB。

关键词: 视频编码, 帧内预测, 纹理一致性, 模板匹配预测, 绝对差和

Abstract: This paper proposes an intra-frame template matching prediction algorithm with boundary texture consistency. Boundary texture consistency constraint is formulated and added to the determination criteria of intra template matching with the purpose of improving the correlation between the generated prediction and current block’s neighborhood. Experimental results show that the enhanced template matching prediction based on boundary texture consistency outperforms intra-frame template matching prediction by an average of 0.11 dB and maximum of 0.17 dB. The new intra-frame coding algorithm can improve the coding efficiency of intra-frame template matching prediction and have good potentials in next generation video coding standard formulation.

Key words: video coding, intra-frame prediction, texture consistency, template matching prediction, sum of absolute difference

中图分类号: 

• TP311