Compressed video QoE assessment based on codec agnostic semantic video model

The amount of multimedia information generated in today society is growing exponentially. This makes an important purpose of research to evaluate the quality of service in order to achieve customer satisfaction, i.e. in order to improve the quality of experience. In literature there are different metrics for assessing the quality of experience and many of these work only on decoded video. This characteristic could be a problem because nowadays videos are transmitted in different encoded forms, such as H.262, H.263, MPEG-4/H.264 or H.261. Hence metrics for evaluating the video quality directly in the compressed domain have been introduced, with the advantage of a low computational complexity but with the disadvantage of coupling between implementation and video standard syntax. In this paper a method based on a "generic encoded video model" is proposed for compressed video QoE assessment in order to evaluate QoE metrics in the encoded domain independently by target codec. This has been achieved by a video stream representation based on a semantic abstraction of the video syntax. In particular we present method for objective estimation of video quality suited for video conferencing and video surveillance on switched networks, where the user interest is mainly focused on the moving objects in the foreground.