MODEL-BASED VIDEO SUMMARIZATION FOR MOBILE USERS

摘要

The use of video information has emerged greatly nowadays due to the availability of acquisition, processing and editing tools using current hardware and software systems. However, we still face some problems to accommodate such data format to the end users. We have to cope with the current limitations of resource capabilities over the distributed architectures and the features of user's devices as well. The efficient and intelligent integration use of video and audio analysis techniques could help to provide alternatives for end users. Thus for example, a high-level video modeling mechanism is required to intelligently parse, analyze and extract the significant audio and video information from digital video libraries. This system would be used to generate an abstraction, which represents a short summary of the original video data. An adaptation stage is also required to react to the user environment's status and configuration. In this paper, we will introduce a system design for video modeling for the purpose of media abstraction. We believe that each domain characteristics are different. Thus, we may need to encapsulate some schema, which contains facts, heuristics and human interests in this domain. We will provide an example of applying this system in sports especially in soccer domain. This work is part of the Mobile Agents Alliance project among university of Ottawa, National Research Council (NRC) and Mitel Corporation in Canada.