The Research Of Video Transcoder And Its Development

As the on going development of multimedia communications and video compression technologies, there are more and more video coding standards. As digital video applications become more popular, there are increasing needs to dynamically convert the video between different formats. A video transcoder is a device that converts video from one compressed format into another compressed format with different bit rate, resolution or syntax. It covers two aspects, one is homogeneous video transcoding, the other is heterogeneous video transcoding. The goal of transcoding research is to utilize the coding information and statistics of the input compressed video stream to facilitate the transcoding process and achieve the best possible quality-complexity tradeoff under given application constraints.The video Transcoder research is a joint project with WWCOMS company. During the research, my major work focuses on the research of motion vector reuse and its related improvement. Moreover, I also finished some testing works about video Transcoder, and got a lot of valuable data which is my necessary evidence of this thesis.In this thesis, I first discussed the principles, the technologies and the realizations between different kinds of video compression frameworks based on correlative video coding theories. I also came up with some improvements in this field.Video coding technologies and their international standards are the base of the video transcoder. In this thesis, I took H.264 as an example to introduce the hybrid-framework of video compression which is the mainstream in current video technologies.Motion vector (MV) reuse is a key technology in video transcoder. And in some situation, especially for video conversion between different standards, using the coded motion vectors efficiently is an important way to optimize the video transcoder. In this thesis, I also explain the importance of motion vector refinement (MVR) which can optimize the motion vectors accordingly and can ultimately improve the transcoded video qualities. Some selected experimental data is also listed in this thesis to prove the theory of motion vector refinement (MVR).Finally, I put forward a video transcoder design based on DSP chip. That is to say, it is a hardware video transcoder which will put all the functionalities of video transcoder to a DSP. This design is cost-effective one and can be easily produced as an multimedia server according.