Research On Content-based Error Control Approaches Over Error-prone Channels

Personal communication is the goal for human communication development, which can fulfill the need of whoever interchange the information with whomever whenever wherever. However, the application environment represented by internet and wireless channel is a kind of time variant and error prone channel, so it becomes an urgent to-be-solved problem of video transmission that how to improve the error-resilience features of video stream against the channel errors, and how to improve the visual quality in the error-prone environment. Traditional error control technologies consider videos as long bit streams, ignoring the underlying semantic information and perform error control in a way as signal processing. From the perspective of video content analysis, this dissertation proposes a content-based error control framework under the error-prone channel, exploring the strong correlation between video content, network resources and video quality. In this way, the author intends to combine the information processing technology and video transmission, aiming at solving the problem of error control in multimedia communication.The dissertation expatiates on the concept framework and architecture of content-based error control methodology, puts forward a content description model, and proposes error control methodologies corresponded to content features of different levels. Moreover, experiments are carried out for validity check and effect evaluation. The primary research and problems solved include the following aspects:! The concept framework and architecture of content-based error control methodology is studied. The framework, characteristics, and classification of this methodology are analyzed. According to the properties of different content features, taking the physiological characteristics of human visual system into account, a hierarchical video content description model is presented. Visual features such as color, density and orientation are in low layer; Motion features are in middle layer; human face is located on the high layer. In this framework, video is segmented into different logical units with salient content feature, and each class of unit is associated with right error control approaches, consequently, good error resilient effect can be obtained.! The error resilient methods to protect human face region is studied. Considering the stability and continuity of face region, and the error propagation effect due to the high error rate in wireless channels, the intra-refresh method is used. The redundant data is not too much owing to the continuity, and the error propagation effect can be stopped accordingly.! Error concealment methods based on motion features are researched. When...