Abstract: Since the latest video coding standard HEVC (high efficiency video coding) uses quad-tree recursive structure to divide coding unit (CU), it adds a lot of computation complexity to the inter-frame prediction and makes coding efficiency decline. This paper proposes a fast inter-frame prediction algorithm based on texture information and stationary regions determination, which is aimed to determine the homogeneity regions and the stationary regions to early skip the CU split. The homogeneity regions and the stationary regions can have accurate motion prediction using large CU size, and needn’t be split into smaller ones, which will reduce the computation complexity. The experimental results show that the proposed algorithm can achieve a reduction of 34% encoding time on average, with 0.038 dB PSNR loss and 0.56% bit rate increase, compared with the original HEVC reference software (HM10.0).

Key words: high efficiency video coding (HEVC), video coding, inter-frame prediction, texture, stationary regions

摘要： 由于最新的视频编码标准HEVC（high efficiency video coding）应用四叉树的递归结构进行编码单元的划分，使得帧间预测的过程极为复杂，编码的时间效率比较低下。针对HEVC帧间预测过程，提出了一种基于纹理信息和稳定区域检测的快速算法，旨在检测出相对平坦的区域和随时间变化较稳定的区域，来提前终止四叉树的编码单元的划分。这些平坦或者稳定的区域用较大的编码预测单元就能进行比较准确的运动估计和预测，不需要再进行更小编码预测单元的划分，因此HEVC编码器的时间复杂度大幅度降低。实验结果表明，改进后的算法编码时间平均节省了34%左右，并且造成的视频质量损失非常低，有0.038 dB的PSNR下降和0.56%的码率增加。

关键词: HEVC, 视频编码, 帧间预测, 纹理, 稳定区域