基于特征建模的H.264到AVS-S2监控视频转码快速帧间模式选择算法研究

针对监控视频从H.264到AVS-S2转码中的高计算复杂度问题进行了研究，提出了一种基于特征建模的快速帧间模式选择算法。首先对少量视频序列通过级联方式转码，根据提取的H.264运动特征和AVS-S2最优模式在线训练建立转码模型，再对后续序列根据模型快速选择转码后的帧间预测候选模式，并基于宏块的前景背景特性进一步减少背景宏块的候选模式数目。实验结果表明算法有效地降低了转码复杂度。由于算法在转码过程中实时训练更新转码模型，因此与传统的模式映射算法相比改善了转码图像质量、降低了转码码率。

A fast inter mode decision algorithm for surveillance video transcoding from H.264 to AVS-S2 based on feature modeling

In order to reduce the high computational complexity for surveillance video transcoding from H.264/AVC to AVS-S2, this paper presents a fast inter mode decision algorithm based on feature modeling. The algorithm firstly transcoded a small number of frames in a cascade transcoder and trained a model based on the motion features in H.264/AVC and the corresponding best modes in AVS-S2. Then the remaining frames were quickly transcoded by selecting the candidate modes for AVS-S2 according to the model. The number of candidate modes can be further reduced for the background macroblocks by judging the foreground or background characteristics. Experimental results show that the proposed algorithm efficiently reduced the computational complexity during transcoding. Due to the fact that this model can be updated in real time according to the video content, this algorithm improved the transcoding performance of better PSNR quality and lower bitrates compared with the previous mode mapping algorithm.