New Techniques Research In Video Transcoding

The development and interactivity of communication network, internet and telecommunication network systems are relatively new and subject to a great deal of research worldwide. The evolution of ASCI technology, CCD technique, video equipments'abilities and screens'resolutions has forced the area of image and video coding to improve at full speed. During 1995 to 2010, ITU-T and MPEG global organizations had drafted more than nine video compression standards. As the diversity of networks, types of devices, and the increasing of video content resolution formats, the interoperability between different systems and different networks is becoming more urgent. Video transcoding technique which born together with video compression standards, is one of the key technologies to solve the mismatch between video devices and networks. This thesis is focus on new techniques of video transcoding researches. The main contributions are as follows:1. Propose a hybrid up-sampling algorithm which combined the advantages both of methods in pixel domain and DCT domain. Experiment results show that the proposed hybrid method can get up to 1dB PSNR gain compared with DCT method and up to 8dB PSNR gain with Wiener filter method. At the same time, the visual quality of the proposed hybrid method is absolutely better than either DCT or Wiener method.2. Design a motion estimation based robust seam curving method. It is one of the attempts to adjust seam curving technique into video resolution adaption area.3. Propose a more accurate rate-control algorithm for video transcoding. The proposed algorithm benefits from the information provided by the input precompressed stream. So it performs better than other rate-control algorithms whether in high bitrate case or low bitrate case and in video streams with small motion vector case or large motion vector case. The proposed algorithm uses the parameters got from the input stream such as QP, frame size, frame rate, R andρto set up a more accurate R~QP model, then calculates a more suitable initial QP value and remaining frames'QP values by considering the importance of the current frame in the whole GOP. Experiment results have shown a noticeable improvement of the proposed rate-control algorithm compared with others.4. Design a video compression via sampling scheme to reduce the bit rate in great deal, and propose a feedback approach for sampling systems'performances improvement. Simulation results display that the proposed method can provide much better visual quality than increasing QP value only scheme.5. Propose a random access location inserting method in SVC and a method to trasncode spatial scalable streams from SVC to H.264/AVC. Experiment results show that the proposed transcoding scheme can get better PNSR than FDR scheme which is the best one in traditional way. The thesis analyzes the reason is that FDR only re-encodes the reconstructed video sequences. Under the rate-distortion restriction, FDR will output a video sequence which looks like the reconstructed squence at best effort. But in fact, the reconstructed video sequence is quite different from the original one. So the re-encoded video may not look like the original one mostly under the same transcoding requirements. However, the proposed method adopts the priori information abstracted from the input streams, which were got from the original video streams and can make the transcoded video sequence look more like the original one. It could be result in PSNR gains at last. Furthermore, the proposed method can save quite a lot of transcoding time.