A robust shot change detection method for content-based retrieval

In this paper, a new approach that combines global and local information to detect shot boundary changes is proposed. For abrupt shot change detection, an adaptive threshold that is applied to histogram differences of a sliding window is used. To detect gradual shot changes such as fades and dissolves, we divide a selected frame into a fixed number of blocks and perform an analysis of block-wise histogram difference. Then, a statistical model imitating the gradual shot transition is proposed. Finally, the proposed algorithm is evaluated and shows good detection results.