Research On Stereo Video Congestion Control Method Based On Macroblock Importance

With the rapid development of network communication and video business, the transportation problems of stereo video have become a research hot spot in the field of network video system. Stereo video has a huge amount of data, it is easy to occur congestion on best-effort network channel. This will not only affect the performance of network system, but also decrease transmission quality of stereo video. Therefore, it is critical to establish a suitable congestion control method. This paper proposes a stereo video congestion control method based on macroblock in a frame has different importance.Firstly, this paper proposes a macroblock level transmission distortion estimation method for stereo video to solve the low accuracy of frame level method and high computational complexity of pixel level method. It builds distortion models for all macroblock modes by analyzing the impact of stereo video data loss on transmission quality. Experimental results show that, it can get the model value approximately equal to actual value under different packet loss rate, while the average distortion estimation error is about 7%, and average distortion fluctuation is about 0.2%.Then, we analysis the impact of the lost macroblock on current frame and the subsequent frames, and distinguish importance of macroblocks with the two parts of effect. The impact of macroblock on the current frame is its transmission distortion, and the impact of macroblock on the follow frames can be obtained by calculating the total number of the macroblock referenced by follow frames. Experimental results show that, the proposed method can get higher Peak Signal Noise Ratio(PSNR) compared with the random packet loss method.Finally, this paper proposes a congestion control method based on macroblock importance to change the limitation of treating the whole frame or sequence equality. It divides macroblocks of a frame into three levels, when detecting network congestion, discard low level of macroblock packets actively according to network congestion degree. Transporting them with Real-time Transport Protocol/Real-time Transport Control Protocol(RTP/RTCP) on NS2, results show that, the proposed method can increase averagely 0.5d B compared with the congestion control method of random packing.