The Research And Implementation Of A Fast Video Transcoder

Recent years, with the fast development of the Internet and mobile network, multimedia application has been widely used. Due to the diversity of the system platform and the terminal, multimedia application usually requires video data that is adapted to a variety of platforms and networks and is also compatible with the new and old equipment. Video transcoding technology can convert a format of the video stream into another kind of video stream. It can solve the problem of cross-platform and cross-terminal compatibility issues. Although video transcoding technology has been studied for many years, with the emergency of massive video data in multimedia cloud platform and gradual application of HD video, the video transcoding technique still faces many challenges. For example, the high computational complexity of video transcoding still restricts the parallel processing capabilities of multimedia gateway. Therefore, investigating a fast and efficient video transcoding algorithm becomes an important research area of video application. It has broad application prospects.This paper analyzed the latest technology and methods in video transcoding area. In order to improve the speed of video transcoding, we investigated the application of pattern recognition and machine learning techniques in it. The main achievements are as follows:First, we analyzed the existing video transcoder based on data mining and proposed an improved Jrip decision tree model. This model improves the problem of complex structure, hard to realize and low efficient of the code of the previous method.Second, a fast macroblock motion vector estimation algorithm based on data mining has been proposed. The proposal can decide the encoded information of motion vector directly. With the motion vector information of target rate, the speed of macroblock motion estimation is raised. The experiment results show that the advanced MPEG-2â†'H.264and H.263â†'H.264transcoder can mostly reduce36%of the transcoding time with a little loss in transcoding quality. Moreover, the transcoding performance is better than the previous method.Thrid, this paper also did some exploration and attempts on the application of deep learning in video transcoding and proposed a fast mode decision method based on DBN. The results show that the DBN network performs well in classification and the video transcoder based on DBN can achieve good video quality. Therefore, it is preliminarily verified the effectiveness of deep learning in video transcoding.