A scale-space filtering-based shot detection algorithm

In this paper we proposed an enhanced graph partition shot detector. In order to increase the detection efficiency, we introduce a scale space filtering-based approach, which makes it possible to derive a relative change ratio measure, instead of an absolute one. Furthermore, in order to reduce the computational complexity, a two-pass analysis approach is proposed. In a first step, the algorithm detects time intervals which can be reliably considered as belonging to the same shots. Abrupt transitions considered as certain are also detected in this stage. In a second step, the analysis is further performed only for uncertain time intervals. The proposed approach shows superior performances with respect to state of the art techniques in terms of both detection efficiency (with a gain of 10% in terms of precision and recall rates) and computational time (in average, 25% faster).