QOS Research For Video Conference

As development of video compression and Wide Band, video conference based on internet has grown quickly, and it has become an important part of our daily life. But the internet does not provide any transmission QOS protection, for its best effort characteristic. Meanwhile, it is still not enough for video transmission though band width has increased. Moreover, the video quality is always affected by the performance of different cameras. The QOS of video conference can not be guaranteed for the reasons above. So, more and more researchers have paid their attentions to how to enhance the quality of video conference. This dissertation aims at the deep research on how to enhance the quality of video conference in the condition of limited band width and normal cameras from the aspect of theoretical and practical.Firstly, this dissertation introduces the research background of video conference and its importance for us. Then, the state of the art of video conference is summarized. In order to deal with the effect of noise to video, an adaptive spatio-temporal video denoising algorithm is proposed. By analyzing the characteristic of space domain and temporal domain, the size of filter can be adaptively adjusted by the estimated strength of noise and the edge information. According to the motion parameters of the encoder of previous frame, the motion parameters of current frame can be estimated, which will be used to select candidate pixels for filter. This algorithm can effectively remove the noise and enhance the video quality.Then the dissertation proposes a novel method to enhance the quality of region of interesting. Face detection and face tracking have been used to obtain the rectangle target area of human face and the background area. In order to increase the quality of the target area, a novel strategy for bits allocation has been adopted to allocate different bit counts to various areas according to their corresponding weights. Moreover, if noise exists in video, different antinoise ways to different areas will be used to reduce the noise.Furthermore, error concealments for packet lost in transmission have been researched. An error concealment mechanism based on encoder and decoder is proposed for whole frame losses. A novel method is used to recover the lost frame in decoder. In order to stop the error propagation as quickly as possible, a strategy of reference adjustment in encoder is adopted by making use of the feedback of the receiver.Finally, scalable video coding (SVC) has been discussed. Based on SVC platform developed by ourselves, the base characteristics of H.264 SVC and the detail of decoding process for layers and macro-blocks have been elaborated, and different methods on realizing the scalable coding in different layers in encoder have been depicted according to our application.