Design And Implementation Of Key Algorithms For H.264Multi-Chnnel Real-Time High-Definition Encoding And Transcoding

With the continuous development of video communication technologyand the terminal equipments, real-time video communication and highimage resolution have become an increasingly demand to people,high-definition (1080P or even higher) has become one of the maindevelopment trends. In the real-time high-definition encoding andtranscoding field, constrainted by the characteristics of the encodingalgorithm and the hardware resources, the encoder cannot use all thecoding tools that the coding standard allows, leading to a substantial gap incoding gain between the actual encoder and the standard encoder. Besides,the single-core embedded platform has been more and more difficult tosolve the contradiction between the high resolution and real-time videoencoding, multi-channel video encoding and transcoding technology basedon multi-core platform has become a hot topic.Video encoding and transcoding tend to focus mainly on two aspects,namely, how to improve the encoding and transcoding speed and try toimprove the image quality as much as possible. These two aspects are oftennot had both, so the research on the encoding and transcoding technologyalso manifests in the two aspects of how to balance speed and quality.This thesis is based on the above background and mainly focus on theinter frame predictor which is the most resource consuming part of theH.264video encoding and transcoding. By analyzing a number of keytechnologies and key algorithms and combining with the advantages of theexisting motion estimation and mode decision algorithms in H.264encoder together with the parallel design strategy of Tilera multi-core processor, weinnovatively propose a fast motion estimation and mode decision algorithmfor multi-channel real-time high-definition encoding and transcoding onTilera Tile-Gx36platform. Besides, we have also proposed a fast motionestimation and mode decision algorithm for H.264bitrate reductiontranscoding. Finally, we optimize and test the algorithms and achieve theH.264multi-channel real-time high-definition encoding and transcoding.The new algorithm proposed in the thesis presents a new algorithmcan effectively reduce the resource consumption of inter frame predictionand increase its processing capacity, as well as enables the encoder to usemore coding tools under the same resource constraints, thus making greatcontributions to further improving the coding gain. Experimental resultsdemonstrate that it can achieve significant complexity reduction withnegligible loss of coding efficiency for high-definition video encoding andtranscoding in Tilera platform. What’s more, our algorithm is still able toachieve the H.264three-channel or more real-time high-definition encodingand transcoding with the use of B frame encoding.