VLBV 98: Oct 8-9, 1998, University of Illinois at Champaign/Urbana
Where are the "Killer-Apps" for video content analysis and retrieval technologies
As computer networks improve and digital image and video libraries become more readily available to homes and offices via the Internet, the problem of bandwidth and one's ability to access relevant data become more challenging. However, video compression and content retrieval had been treated as separate problems in the past. Fortunately, this is changing as a result of MPEG-4 and 7 efforts and intensive research in image and video content analysis. Compression for efficient transmission and storage and content-based coding and indexing for efficient access and retrieval finally are addressed under one unified framework, which starting to unite scientists and engineers from multiple research disciplines.
Many compression algorithms and standards have been developed and adopted in the last two decades for efficient video transmission and storage, thanks to intensive researches and standard activities. In the last few years, visual content analysis and content-based image and video indexing and retrieval research have also achieved significant progress and generated many algorithms and system prototypes. However, when we look at the impact of image compression research and that of image and video content analysis, we see a huge difference: there are very few convincing stories one can tell about successful applications of video/image content analysis and indexing.
Due to limitations of low-level feature-based image content representations and incapability of automated visual content understanding, visual content-based retrieval of images and video clip is far less than satisfactory and often unpredictable. Therefore, the applications of low-level feature-based image/video indexing technologies are limited to very specific application domains, where search of images relies on color and texture patterns which are hard to describe using text. Market of such applications can be found in some professional market such as images and video stock footage houses.
Partitioning video into shots provides a foundation for organizing video clips according to their temporal structures. This can be further enhanced by grouping shots into scenes or groups of similar contents, based on low-level features and often domain knowledge. Abstraction of video into key-frames and highlights reduces the data amount for video handling and browsing, while maintaining important content information. These video processing technologies have found applications in news and stock footage video archiving, and provide some enabling tools to fast and content-based video browsing.
However, the market scale of image and video content analysis and indexing technologies remains small. It was blamed that video capturing, digitization and compression tools were too expensive. This blame can no longer hold as video camcorders become a house hold appliance item and the price of video digitization and compression devices is falling into $199 to 399. Many researchers, especially those in industrial research labs, have started thinking harder where is the "killer applications" of the video content analysis and indexing technologies, while they continue developing new and improving current technologies. Some are more pessimistic about the current video content analysis and indexing technologies and their development potential: they offer too little to make any significant impact, especially when compared with compression technology. Will there be a major impact of video content analysis and indexing technologies? Are we falling into the same hype that computer vision once did? If not, where are the killer applications of the technologies? These are the hard questions we have to answer.
This panel consists of expert and pioneers in video content analysis and object-based video coding, and they will share their view on where the killer applications of the technologies which they have been leading the development in the last decade. The application areas to be addressed include home appliances, TV production and transmission, multimedia databases, and the Internet.