Studies On Compressed Digital Video Superimposing Based DM642 Platform

Appearing and developing of video compression technologies are of great significance, which enable the storage and exchange of programs in compressed format. With MPEG-2 programs increasing, the research on compressed video became a hotspot.Video superimposing is widely used in DTV implement, which includes both TV logo insertion and caption adding. TV station or other TV content manufacturer always deals it with caption machine. But traditional caption machine have some specific defeats. Firstly, nonlinear editing is based on fully decoding and encoding, which will lower the video quality, although it can fulfill many complicated effects. Secondly, traditional caption machine is based on X86 architecture, and most of local caption machine run Windows as their OS, which will make caption machine unstable and unsuitable for long time-span usage.The main task of this paper was designed digital television captions superimposing system based on DSP platform. Making full use of the advantages of DSP with rational deployment DSP resource and optimization of MPEG-2 encoding and decoding process focus on caption or logo area minimize the impact of the original video quality. DSP-based platform itself has the characteristics of stability, low power consumption and so on.Firstly, the principle and basic knowledge of MPEG-2 is introduced. the current implemention of logo insertion and caption adding is also discussed. Besides, we analyse the current situation of TV logo insertion and caption adding system, like caption machine, etc. Secondly, this thesis design and implement the DSP framework of Digital TV stream caption superimposing. Choose the mothed partly-decoding then caption adding then partly-encoding, which caption operation is in spatial domain of video.Thirdly,in order to assure video quality would not be lowered by the operation of logo insertion, we use the method of partly-decoding then caption adding then partly-encoding. Then the area where the macro blocks should be decoded is the key. On the base of analysis and comparation among all the available algorithms, the enhanced"backtracking"approach is proposed, which can make motion estimation an easy job in the latter reencoding process. Compared with the original"backtracking"approach, the enhanced one will result in a less changed stream in bit rate.Lastly, the entire logo insertion and caption adding system is build on a TI DSP chip with the name of TMS320DM64. We manage to optimize program to DSP platform in many ways, from top to bottem, which refine the code under the principle of DSP and the characteristics of TMS320DM64, but also use TMS320DM64's resouces reasonably and flexibly. As the approach partly-decoding then caption adding then partly-encoding is involved with MPEG-2 decoding and encoding itself. The quality of the logo insertion and caption adding is pretty good, and the system efficience meets the real-time speed need on actual implement.