DCT域HDTV至SDTV视频转码器关键算法的研究与实现

【作者】 梁铁；

【导师】 朱维乐；

【作者基本信息】 电子科技大学 ， 信号与信息处理， 2006， 硕士

【摘要】 DCT域HDTV至SDTV视频转码,与传统的像素域转码相比,不仅避免了DCT、IDCT运算,还通过利用输入HDTV码流中宏块的运动矢量信息直接构造转码后SDTV相应宏块的运动矢量,避免了运算量极大的运动估计过程,有利于HDTV至SDTV实时转码的实现。本文主要研究在DCT域内基于宏块公共信息的运动补偿快速算法和利用DCT系数进行码率控制算法。本文给出的运动补偿快速算法与传统的DCT域运动补偿算法相比,可以节省44%的运算量。给出的码率控制算法与传统的TM5算法相比,不仅使编码生成的图像信噪比有所提高,而且解决了场景切换问题。转码的关键算法由C语言实现,并通过基于Intel Pentium 4 CPU具备SIMD功能的SSE2汇编指令集优化运算量大的转码程序,最终在PC平台上实现了在DCT域内HDTV至SDTV的快速视频转码。更多还原

【Abstract】 Compared with the traditional pixel domain video transcoding, the transcoding from HDTV to SDTV in the DCT domain, not only avoids the DCT and IDCT transforms, but also avoids motion estimation, the most computation costing process, by making full use of the macroblock motion vector information from the input HDTV stream to construct the motion vector for the according SDTV macroblocks. It’s helpful to achieve the HDTV to SDTV video transcoding in the real time.This dissertation presents the research on the fast motion compensation algorithm with the shared macroblock information in the dct domain and rate control algorithm with dct coefficients. The fast motion compensation can save 44% computation compared to the traditional motion compensation in the dct domain. Compared with the TM5 method, the rate control algorithm can not only increase the encoding images PSNR, but also solve the scene change problem.The key transcoding algorithms are achieved by the C language, and we optimize the most computation costing transcoding programs with the SSE2 assembly language, which is based on the Intel Pentium 4 CPU and has the SIMD function. At last, we achieve the quick video transcoding from HDTV to SDTV in the dct domain on the PC.更多还原