Research On Key Technologies Of Video Broadcast Quality Service Combined With The Dynamic Motion Characteristics

With the rapid development of Internet and multimedia, video streaming media data is now growing explosively. This leads to many new challenges on video streaming media processing and transmissing. On one hand, people demand more and more multimedia resources and higher and higher video quality on the receiving end.This requires the provision of data storage and transmission capacities matching streaming media resources.On the other hand, due to the restrictions of computer processing power, heterogeneous network environment, network bandwidth, terminal processing capablity and so on, people can only assign limited computational resources to important visual information analysis and processing. The network development trend is digital broadcasting, wireless network, broadband communication and high density storage. Under this trend, efficient video encoding, heterogeneous network transmission and error recovery capability should be considered simultaneously. These technologies can effectively solve the contradiction between the network bandwidth, video quality and real-time user access. Also the quality of streaming media broadcast communication can be improved while reducing costs. In order to accurately assess the quality of the video broadcast, objective video quality assessment is needed. This video quality assessment can provide the basis for video processing and transmission system to improve the human visual experience.Based on the above analysis, this paper focuses on the key transmission technologies in order to improve video quality with H.264/SVC precoded video stream. The research is based on analysis of human visual perception characteristics. The involved key technologies include: visual salient region extraction, packet scheduling policy, video stream progressive transmission, network congestion control mechanisms and video quality assessment. The content and innovation can be summarized as follows:1. The video dynamic motion characteristics description method reflecting human visual perception is proposed. Combined with the characteristics of human visual perception and eye movement data analysis, this method introduces motion intensity, motion complexity and motion density.The weighted sum of the front motion vectors and background ones is proposed to describe the dynamic motion characteristics of the corresponding frame. The obtained object motion value can ccurately reflect human visual perception attention.2.The salient region extraction method consistent with human visual characteristics is proposed. The anlysis of log spectra in frequency domain is used to obtain spatial saliency map. The temporal saliency map is extracted according to dynamic motion characteristics description and visual attention model. Then spatial and temporal saliency map are weighted accumulated by dynamically adjusting the weight. The final extracted salient regions are consistent with human visual perception.3.The packet scheduling policy based on video dynamic motion characteristics is proposed. On the basis of coding structure analysis and rate distortion model, this method integrates video dynamic motion characteristics description and takes into account the objective and subjective factors affecting video broadcast quality. First the importance of packet is analyzed and video stream is sorted from high to low according to the importance degree. Then the packetes in the sending buffer queue are scheduled under the constraints of bandwidth and delay. The packetes with the higher level of importance are transmitted prioritily. By this method, the video broadcast quality is improved significantly.4.The network congestion control mechanism for video stream transmission is proposed. On the basis of the widely used active queue management algorithm ARED, this algorithm modifies the calculation method of average queue length in the buffer and maximum drop rate. When discarding packets, the packets with lower priority are discarded firstly. Not only the network performance is effectively upgraded, but also the human visual perception is improved significantly.5. The object video quality assessment is proposed. Based on the structural similarity index measurement system, the quality evaluation value of each frame is weighted accumulated by saliency. Then the weighted sum of every frame quality assessment value is calculated according to video dynamic motion characteristics description. The experiment results on public data sets and testing of key technologies metioned in this paper show that this evaluation index can better reflect the subjective human visual perception and is more consistent with the subjective evaluation results.In summary, for the purpose of improving video broadcast quality at the receiving end, this paper specifically studies on the key technologies involved in the video transmission. The research is mainly based on video dynamic motion characteristics description. A large number of experimental result anlysis shows that the key technologies of video streaming broadcast quality service studied in this paper can significantly improve the subjective human visual perception.