Abstract :
Automatic video partitioning is the first step for the content-based parsing and indexing of video data. Many methods have been introduced to address this problem, e.g. pixel-by-pixel comparisons and histogram comparisons. Histograms are robust to object motion, and therefore they are used widely, but they are sensitive to the illumination variation that is inherent in the video production process, especially in TV news reports, so false alarms are inevitable. The color ratio histogram, which is robust to illumination changes, has been proposed as a frame content measure to solve this problem. However, it is computationally expensive. In this paper, efficient technologies are proposed to discard false alarms that are due to illumination variation. Abrupt lighting condition changes last a very short time, and the content of the image will return to what it was before the change. Therefore, these changes can be recognized and discarded by comparing the histograms of certain frames. Another illumination-invariant measure, called ABEM (Area of Boundary in Edge Map), is used to deal with gradual illumination variations. Compared with the color ratio histogram, ABEM is more economical regarding the computational cost while maintaining a good performance
Keywords :
database indexing; graphs; lighting; video databases; video signal processing; ABEM; TV news reports; abrupt lighting condition changes; automatic video partitioning; color ratio histogram; computational cost; content-based parsing; edge map boundary area; false alarm avoidance; frame content measure; gradual illumination variation; histogram comparisons; illumination-invariant measure; image content; object motion robustness; performance; pixel-by-pixel comparisons; video data indexing; video production process; video shot detection; Automation; Cameras; Gunshot detection systems; Histograms; Indexing; Lighting; Pattern recognition; Robustness; Streaming media; Video on demand;
