Graphics: Visual and Interactive Computing
Video Visualization
Please also visit the Swansea volume graphics gallery. |
What is the Need? The rapid advance of digital technologies has resulted in an explosion of digital imagery data. In particular, video data, generated by the entertainment industry, security and traffic cameras, video conferencing systems, digital camcorder and not mentioning video emails, internet videos, etc., is making their ways into everyday life. It is almost certain that there will be a multi-fold increase in video data in the coming years. A video is a piece of ordered sequential data, and viewing videos is a time-consuming and resource-consuming process. For example, an increasing problem in the security industry is the ratio of surveillance cameras to security personnel. Imagine that security officers have to review an overnight collection of video tapes when they arrive at their desks in the morning. It is simply not possible for any security officer to study a large number of video tapes everyday. It is hence highly desirable to develop methods for extracting and highlighting interesting features in video sequences. What about Vision and Statistical Methods? Automated video processing is a research topic that is of significant importance to the security and entertainment industries. There is a rich collection of techniques for analyzing imagery data, and for computing various statistical indicators. However, there is a general lack of effective techniques to convey complex statistical information intuitively to a layperson such as a security officer, except using line graphs to depict 1D signal levels. Most of the techniques have not reached such an intelligent level that they can be relied upon to make decisions in place of a human. What is Video Visualisation? Video visualisation is a computation process that extracts meaningful information from original video data sets and conveys extracted information to users in appropriate visual representations. We developed a novel method for handling large volumes of video data, and proposed to employ volume visualization techniques for "summarising" video sequences, and rendering video volumes into appropriate visual representations that can be used to assist in the decision making processes of a human operator. For example, every morning, security officers can be presented with one or a few visualizations for each surveillance camera that has been monitoring a premise during the previous night. From the visualizations, the officer can observe the levels and patterns of the activities recorded. We believe that such visualizations can convey much more information, especially spatial information, than a few statistical indicators or line graphs. With carefully prepared visualizations, the human vision system, perhaps the most intelligent vision system, is able to become accustomed to certain kinds of "normal" visual patterns, and react to unusual levels or patterns of activities that need further investigation. Video visualization can also be used to assist in processing videos, such as video segmentation, and video annotation. Three Hypotheses. Considering video visualization as a new scientific subject, we propose the following three hypotheses:
The Main Contributions of our Recent Paper in IEEE Visualization 2003. As no previous work on this subject was found in the literature, our paper, entitled "video visualization", serves as a brave path-finder. Our main contributions include:
For the third hypothesis, we believe that some comprehensive user studies are necessary, but we are confident about the likelihood of a positive conclusion. Benchmark Resources. The following video clips can be used as benchmark problems for further development in this area:
Main References
|
