节点文献

转码技术中H.264模式决策快速算法的研究与实现

Research and Implementation of Fast Mode Decision in to-H.264 Transcoding

【作者】 马妍;

【导师】 陈纯; 张海翔;

【作者基本信息】 浙江大学 , 计算机应用技术, 2006, 硕士

【摘要】 随着数字视频及多媒体技术的发展,适用于不同领域的MPEG,H.26x等国际视频编码标准应运而生。为了适应用户终端的多样性以及网络自身的传输特性,常需将已压缩的视频码流进行转换编码,以实现资源共享。新一代视频标准H.264/AVC具有当前最高的编码效率。因此从其它视频编码标准转换到H.264很有实际应用价值。但H.264使用了一系列计算复杂度很高的新算法,使得这一转换过程相当费时。快速转码算法的研究已成为当前的研究热点,快速宏块模式决策正是其中的关键技术之一。 介绍了H.264的编码特点,并与以前的标准进行比较。回顾了转码技术现状以及转码过程中的关键性问题。分析了典型的帧内和帧间模式决策算法,并指出了这些算法的不足之处。分别提出了编码转换过程中H.264的帧内和帧间模式决策快速算法。 在帧内预测模式选择上,利用了变换域内系数直接转换的的优势,节省了IDCT和HT两个复杂的计算步骤:使用了两个复杂度相差较大但又高度相关的代价函数依次对候选模式进行筛选,既有效保留了最优帧内测模式,又降低了模式决策的计算复杂度。 在帧间预测模式选择上,充分考虑到了H.264的多宏块模式和多参考帧等特点,提出了基于多帧参考的帧间快速模式决策算法。将H.264中的宏块模式加以分类,减小了映射模式的选择范围。自适应的运动估计搜索范围确定方法,可以快速有效的合成新的运动矢量。结合提前退出和由上至下宏块分割等方法,加快了帧间模式选择。 所提出的帧内和帧间模式决策算法适用于多种编码标准向H.264的转换,具有较好的通用性。在篇末的转码系统中实现了以H.263、MPEG-2、MPEG-4简单档次编码的视频向H.264的转换。

【Abstract】 With the development of digital video and multimedia technology, the series of MPEG and H.26x video coding standards have been widely used in different applications. In order to adapt to the diversity of terminal users and features of various networks, a compressed video bitstream is often transcoded from one format to another and shared by different users. The emergence of the new generation video coding standard H.264/AVC greatly contributes to the video’s compression and transmission. The to-H.264 transcoding is becoming an active research topic and fast mode decision is one of the key techniques in the process of to-H.264 transcoding.In this paper, the status quo and key problems in transcoding technology is presented. The features of H.264 is analyzed and well compared with other coding standards. After going over several typical mode decision algorithms, novel fast mode decision methods are proposed considering two aspects, intra-frame and inter-frame prediction, when transcoding one format of video sequence to H.264 format.For intra-frame prediction, taking the advantage of direct coefficients conversion in the transform domain, IDCT and HT are saved. Two computationally different but highly correlated cost functions are employed as a criterion to choose the optimal intra-frame modes.For inter-frame prediction, the multiple modes and multi-frame reference features of H.264 are well considered to efficiently decide the block size to use. The proposed adaptive motion estimation search range, early-stop and top-down splitting methods contribute to high-speed transcoding with negligible PSNR drop and little bitrate increase.The proposed algorithms can be applied to transcoding several video formats to H.264. The system implemented at the end of this paper can transcode video encoded by H.263, MPEG-2, MPEG-4 simple profile to H.264.

  • 【网络出版投稿人】 浙江大学
  • 【网络出版年期】2006年 09期
  • 【分类号】TP391.41
  • 【被引频次】1
  • 【下载频次】158
节点文献中: