Abstract: In order to reduce the computational complexity of macroblock（MB） mode decision for B frame in AVS（Audio Video Coding Standard），a fast strategy for B frame MB mode decision based on RDO（Rate-Distortion Optimization） technique is proposed.According to the characteristics of eyes which are sensitive to edge，a visual threshold mode is proposed.Edge information of an MB is extracted by sobel operator，which is then applied to visual threshold mode to get the edge modes of an MB.Edge mode is chosen based on the analysis of the grain complexity of one MB from both the horizontal and vertical directions，in order to get fast strategy of MB partition mode decision.And the intra mode decision algorithm is proposed based on the temporal and spatial correlation.Experimental results show that the new algorithm reduces computational time by 40%，with slight increment of bit-rate and little loss of peak signal-to-noise ratio.

Key words: Audio Video Coding Standard（AVS）, macroblock partition for B frame, Rate Distortion Optimization（RDO）, grain complexity of video sequence, visual threshold mode, sobel

摘要： 为了降低AVS中B帧宏块模式选择的计算复杂度，提出了一种基于率失真优化（RDO）的宏块模式快速选择算法。根据人眼对图像边缘比较敏感的特点，提出了视觉阈值模型；利用sobel算子提取出宏块的边缘信息，并应用到视觉阈值模型中，从而得到了边缘点模型。在对图像水平、垂直纹理复杂度分析的基础上，对边缘点模型进行了选择，得到了宏块分割模式的快速算法；根据图像序列的时/空相关性提出了帧内预测判决算法。仿真结果表明，与原AVS全搜索算法相比，本算法在比特率略有增加和信噪比基本保持不变的条件下，使整个图像的总编码时间减少了40%左右，从而大大提高了AVS编码速度。

关键词: 音视频编解码技术标准（AVS）, B帧宏块分割, 率失真优化（RDO）, 图像纹理复杂度, 视觉阈值模型, sobel