Research On Error Concealment And Quality Assessment Model For High Definition Video Under Packet-loss

Recent years, with the development of computer processing capability and network technology, a number of newly video applications are rapidly arising, such as video telephony, video conferencing, IPTV and digital TV. Video compressing technology plays the key role in pushing forward the process of all these applications. However, compressed bitstreams are very sensitive to channel errors, for the adoption of high efficient Variable Length Coding and predictive coding technologies. Impairment due to packet loss will greatly degrade the video quality at the decoder. Therefore, quality assessment for video under packet loss is very important for related applications.Traditional quality assessment models for video under packet loss neither consider the practical characteristics of the decoder, nor take the content of video sequences into account. Furthermore, current video assessment model are usually developed for low definition videos. With the rapid spread of high definition video (HDV) applications, it is a must to develop specified quality assessment model for HDV. In this paper, three classical temporal error concealment (TEC) algorithms are compared aimed at the HDV. Through the comparison of Mean Opinion Score (MOS) and Peer Signal to Noise Ratio (PSNR), Decoder Motion-Vector Estimation (DMVE) algorithm was chosen as the proper error concealment for HDV. Moreover, it is proposed in this thesis that adding a linear filter after the error concealment process would help to eliminate the blocking artifacts.Subjective assessment experiments demonstrated that video content is one of the key factors which affect subjective quality after error concealment, and video content can be expressed as the average of maximum motion vectors, variance of Macroblocks and colour-depth. It is noteworthy that it is the first time for colour-depth to be proposed for quality assessment. Based on the above analysis, a quality assessment model for HDV under packet loss was proposed. Using the nonlinear least square method, the coefficients of this model have been obtained, considering the effect of both packet loss and video content on subjective quality. Experimental results confirm that the predicted quality using the proposed model has good consistency with the human visual system.