| With the rapid development of digital video techniques, many international videocompression standards have been issued. Digital video techniques have been used inmany applications. Video transcoding is a new research filed, it re-code the coded videobitstreams, and convert the coded video to the format, which meets the variousapplication requirements. As a key technique, transcoding techniques will be widelyused in the Universal Multimedia Access (UMA) systems. Demands for transcodingcome from following aspects: 1. Several video compression standards have beenadopted in different industry fields. In order to share the video resource, transcoder isused to convert the syntax and coding tools. 2. The real transmitting channel is instable,video bitrate should be adjusted to the available bandwidth. 3. Various limitationsincluding store capacity, display resolution and computation ability may exist indifferent multimedia terminals;video data should be transcoded according to therequirements of terminals. 4. Special usages such as video encrypting anderror-resilience enhancing.Some transcoding algorithms have been reported recently. This paper first review andanalyze many transcoding architectures and techniques, then give a summary on thetranscoding research. Base on the analysis, some new transcoding algorithms are alsoprovided in this paper. In order to evaluate the performance of different algorithms, asoftware transcoder which support many video compression standards is designed.MPEG-2, H.263 and MPEG-4 are the most popular video compression standards,heterogeneous transcoding techniques between these standards are investigated in thispaper. The difference of these standards are analyzed, detailed transcoding methods aresuggested, such as motion vector (MV) reuse and macroblock(MB) type conversion. Inorder to meet the field to frame conversion demand of MPEG-2 to H.263 and MPEG-4transcoding, this paper proposed a DCT coefficients based field to frame MVcomposition method. Chroma error drift may occur in heterogeneous transcoding. Thedrift reason is analyzed in this paper, and a chroma error compensating transcoder isproposed.Logo insertion of transcoding is highly desired by many applications, deeply research ofthis technique is fulfilled in this paper. A fast luminance logo inserting transcoder issuggested. In addition, MV and MB type optimization algorithm are also provided.Rate control is the most important algorithm of video compression, virtual buffer is usedas feedback in many rate control methods. Based on the distribution of DCT coefficients,a new rate control idea is proposed in this paper. DCT coefficients are stored in thedecoding part of transcoder, and the distribution is modeled. Then, a mathematicrelation between quant parameter and coding bits is established. The mathematicrelation can be used to select quant parameter for individual MB adaptivly, good ratecontrol performance is achieved.Information theory and rate-distortion theory are the base of video compression. A briefreview of related knowledge is given in this paper, then the distortion of transcoding isanalyzed. Rate-distortion theory is very complicated, it can be used as a guide fortranscoding strategy. Some primary research is given in this paper. Rate-distortion basedbitrate and spatial resolution transcoding architecture is suggested, optimizing strategiesare also given.In summary, this paper gives a general research on video transcoding. Some newtranscoding architectures and algorithms are proposed, experimental results showed thatthese methods not only achieve good transcoding performance, but also have somepractical usages. All the research work is realized in a software transcoder. Thetranscoder is designed with C language, modular design method is applied, whichenable easy debug and extension..
|
PDF Full Text Request