一种快速的DCT域MPEG-2到MPEG-1下采样转码技术

A Fast Transcoding Technique From MPEG-2 To MPEG-1 in the DCT Domain with Downsampling

文章提出了一种快速的DCT域MPEG-2到MPEG-1准卷积下采样算法。在转码过程中,头信息保持不变,且要求将每四个相邻MPEG-2宏块变为一个下采样MPEG-1宏块:用最大最小距离方法确定下采样宏块的运动向量、用简单多数原则确定下采样宏块类型以及用加全平均方案确定下采样宏块的量化参数。另外,对下采样视频转码失真来源进行了分析。实验结果表明我们提出的转码方案,在同样条件下与级联像素域转码器(TM5)相比,不仅其计算复杂性减少67.6%、PNSR提高0.1dB,而且具有很小的比特控制错误。

In this papar,we proposed a fast transcoding algorithm from MPEG-2 to MPEG-1 with 2：1 quasi-convuluted downsampling in the DCT domain. During MPEG-1 transcoding processing, the head information obtained from the incoming MPEG-2 video stream are kept unchanged, The motion vector, macroblock type, quantiztion parameter are directly obtained from the corresponding four macroblocks available in the incoming MPEG-2 bitstream information according to maximizing and minimizing distance,the simple Majority rule and quantiztion parameter weighted average respectively. In addition, the source of transcoding distortion for reduced spatial resolution reduction transcoding was analyzed.The experimental results show that our method not only achieves decrease by 67.6 per cent in the computational complexity and and improvement with 0.1dB peak signal-to-noise ratio（PNSR） but also yields a much smaller bit control error, when compared with the Cascaded Pixel-Domain Transeoding Methods with TM5 rate control algorithm.