Download now Free registration required
This paper proposes a new framework of visual mining of multimedia data. The key idea is to integrate data visualization and data mining. Based on this idea, it has developed a prototype system with several critical features to facilitate knowledge discovery. First, it decomposes and represents multimedia data as a set of continuous variables and event variables. Second, it developed various ways to visualize these two kinds of variables separately and together. Third, it visualizes not only raw multimedia data, but also all intermediate and final results of data mining, which allows researchers to access the "Ground Truth" of an experiment along with the results. Fourth, it provides a flexible interface between visualization tool and data mining tools users may use.
- Format: PDF
- Size: 1713.6 KB