Research On Error Resilience Techniques For Video Communication Over IP Network

AbstractAs the crucial component of multimedia communications, video has high demands of networkbandwidth, latency and loss in terms of real-time transmission. However, the best effort orreliability of data transmission provided by the current IP network doesn't offer any quality ofservice (QoS) guarantees for video transmission. In addition, the heterogeneity of the networksand end-systems make it difficult to transmit video in an efficient and flexible way. It isinevitable for a communication system to generate transmission errors due to variable networkcharacteristics, which will result in decreasing of video reconstructed quality. Furthermore, errorswill propagate in temporal and spatial domains, which will seriously deteriorate subjectiveperceptual quality. In this paper, the solutions to error resilience are presented from theperspectives of network and end system. The goal of research is to relieve the negative influencecaused by errors and improve the reliability and efficiency of video transmission.Two novel methods of Intra update are proposed in this paper. The first one defines a parametercalled Sensitivity Degree, which combines the properties of video service and capability of errorrecovery at the receiver side. The addressed Intra update strategy is discussed based on theparameter. Another one gives a more efficient update method based on the feedback informationof the communication channel. These two methods overcome the disadvantage of low codingefficiency while keeping the features of self-recovery to errors in traditional periodic Intra updatemethod.Shannon's separate coding theorem is not applicable in video communication. In the next sectionof the paper, joint source channel coding scheme is employed on the basis of features of sourceand channel coding. A simple source distortion model is introduced to estimate the distortion ofsource coding, and the packet-based channel coding is applied to estimate the end-to-end channeldistortion. Two joint source channel coding strategies are proposed based on H.263 source codingand RS channel coding respectively.In the consideration of the effect of network behaviors, the traditional QoS mechanism isanalyzed and a new video-oriented QoS method which applies the scalable coding of H.263 isproposed. Then a service-based QoS qualitative description method and a QoS negotiationmechanism are proposed. Congestion control is implemented at both transport layer and networklayer using queue management based on the importance of data and rate control method of theendpoint.A test bed for video communication over IP network has been established to verify the feasibilityof the proposed QoS mechanism and error control methods.