Design And Implementation Of H.264 Bitrate/Resolution Downscaling Transcoding

Multimedia communication has become an important style of people's daily communication. Image and video, which can bring people most vivid visual expierence, are being widely concerned and applied. For different application, video streams with different bitrate, different resolution and different formats are generated using different video compression methods. While in the entire communication system, there exist great differences in teminal processing ability, storage capacity, network bandwidth and infra-stucture. Hence how to achieve seamless connection between different teminals and networks has become a hub of multimedia communications research. Video transcoding can convert video streams in bitrate, spatial resolution, temporal resolution and format and provides effective solutions for the achievement of seamless connection between heterogenous networks, different access equipments as well as different compression formats. Therefore, there exists very high value in the research of video transcoding.This paper introduces the practice and research work in H.264 bitrate and resolution downscaling transcoding.First, the characteristics and performance of different type of framework of video transcoding is studied. According to the target of the research work of this paper, a high-efficiency cascading pixel domain transcoding framework is given as the basic frame of H.264 bitrate and resolution downscaling transcoding.Secondly, an H.264 decoder in C language that comforms to H.264 baseline profile and main profile is implemented. Necessary improvement of codes and algorithms is done in purpose of high performance the further designed transcoder. As part of high-efficiency concatenated transcoder, the function and relation between different modules of the decoder are discreetly considered in order to provide good flexibility.Thirdly, JVT-G012 recommandation for H.264 rate control is adopted and reasonably modified to realize accurately rate control in transcoding. On designing the transcoding algorithms, we come up with an intra frame mode refinement scheme based the texture of intra mode and a bottom-up block-merging algorithm to reduce the complexity of mode decision. The result shows the processing speed of the designed transcoder is about three times of that of the full transcoder, with the loss in mean PSNR of transcoded video within 0.5 dB.Last, an H.264 transcoder with 2:1 spatial resolution downscaling is realized and a primary mode refinement and mapping algorithms and motion vector re-estimation scheme are applied in this transcoder. In inter frame mode refinement, the bottom-up block-merging method is also adopted. The result shows the processing speed is almost twice of that of the full transcoder, with only 0.1 dB loss in mean PSNR of transcoded video.