论文部分内容阅读
视频编码中帧内预测利用图像自身的信息降低空间冗余,但是复杂的编码单元(CU)划分和预测模式选择给编码带来了巨大的计算复杂度。针对HEVC帧内编码高复杂度的问题,本文基于直方图分析和率失真(RD)代价统计,提出一种快速帧内编码算法。首先统计每个编码树单元(CTU)中四个32x32 CU亮度分量的直方图,根据直方图统计数据与CU尺寸的关系,确定当前CTU编码尺寸的范围;然后在帧内编码过程中,利用相邻空间区域已编码CU的RD代价计算得到RD代价阈值,判断位于编码尺寸范围内的当前CU是否继续划分。实验结果显示,本文提出的算法相对于HEVC标准模型可将编码速度提高40%,而平均比特流损失控制在0.7%左右,同时视频的视觉质量几乎没有任何变化。
Intra-frame prediction in video coding uses the information of the image itself to reduce spatial redundancy, but the complicated coding unit (CU) partitioning and the selection of the prediction mode bring tremendous computational complexity to the coding. Aiming at the high complexity of intra coding in HEVC, this paper proposes a fast intra coding algorithm based on histogram analysis and rate of distortion (RD) cost statistics. Firstly, the histogram of four 32 × 32 CU luma components in each coding tree unit (CTU) is counted, and the range of the current CTU coding size is determined according to the relationship between the histogram statistics and the CU size. Then, in the intra coding process, The RD cost of the CU in the spatial region is calculated to obtain the RD cost threshold, and it is determined whether the current CU located in the coding size range continues to be divided. Experimental results show that the algorithm proposed in this paper can improve the encoding speed by 40% compared with the HEVC standard model, while the average bit-stream loss control is about 0.7%. At the same time, the visual quality of the video has almost no change.