Robust video shot change detection

In th i s paper we present a novel improvement to existing schemes for a b r u p t shot change detection. Existing schemes declare a shot change whenever t h e f rame to f rame histogram difference (FFD) value is above a particular threshold. In such a n approach, a high value for t h e threshold results in a small number of false a larms a n d a large number of missed detections while a low value for t h e threshold decreases t h e number of missed detections at t h e expense of increasing t h e false alarms. We a t t r ibu te this si tuation to t h e fact t h a t t h e FFD cannot be reliably used as t h e sole indicator for t h e presence of a shot change. In t h e proposed method a two step shot detection s t ra tegy is used which selectively uses a likelihood ratio (computed directly f rom t h e frames and not f rom the histograms) t o confirm t h e presence of a shot change. Such a two-step checking increases t h e probability of detection without increasing t h e probability of false alarm. T h e improvement proposed is simple and computationally cheap. Tests with a wide variety of video sequences prove t h e efficacy of t h e proposed approach.