Evaluation Of The Effect Of Packet Loss On Frame Quality

To provide high quality video services for users over time-varying networks,it is necessary to monitor the quality of the video at network nodes in real-time.Packet loss is a crucial factor that significantly affects video quality.Therefore it is necessary to evaluate the distortion caused by packet loss for video quality assessment.Since a video is composed of sequential frames,an important issue is to evaluate the effect of packet loss on the quality of individual frames.In this thesis,this effect is evaluated according to the characteristics of networked videos using the parametric bit-stream-layer model.Extensive subjective experiments have been carried out to evaluate the effect of the temporal complexity of a video on the frame quality,based on which a metric to measure the temporal complexity is proposed for networked video applications.Firstly,the frame distortion directly due to packet loss and due to error propagation is assessed respectively,based on the proposed metric of temporal complexity,considering the situation of only one frame lost in a GOP(Group of Picture).Two models have been proposed to assess the frame distortion directly due to packet loss and due to error propagation,respectively.Secondly,the joint effect of the frame distortion directly due to packet loss and due to error propagation is studied,where two frames loss is considered.Based on the related observations,an object model is proposed to assess the frame quality which is affected directly by packet loss and by error propagation simultaneity.Thirdly,the frame distortion due to error propagation is researched,where two frames loss is considered.The results show that the object model to assess the frame distortion due to error propagation in which there is only one frame lost in a GOP is suitable for the condition of two frames lost.Finally,an object model to assess the quality of the frame of the networked video with packet loss is proposed based on the above research.Experiment results suggest that the proposed model is valid in assessing the impact of packet loss on the frame quality in a network video.