Improvement of shot detection methods based on dynamic threshold selection

Currently, most shot detection methods proposed in the literature are based on well-chosen static thresholds on which the quality of results largely depends. In this paper, we present a method for dynamic threshold selection (DT) based on clustering a set of N points on a comparison curve, which we use for characteristic feature comparison through images in a video sequence to detect shots. In this method we recursively chose N successive values from the curve. Then by using the clustering method on than, we partition this set into two parts, larger values in El, and smaller values in E2. We try to modelize the form of the curve as a bimodal one, and try to find a threshold around a valley area. Using above clustering analysis, we first apply Color Histogram (CH) and Double Hough Transformation (DHT) that we reported in our previous work on 90 minutes of video sequence. The experimental results show that dynamic threshold based methods improve the static threshold based ones, reducing false and missed detection, and that dynamic threshold based DEFT is more robust than dynamic threshold based CH. Besides, further analysis of 3D indices and lines extracted by DHT through the video sequence allows to detect special camera effects like zoom in, zoom out and camera panning, and gives us different motion vectors through the video sequence.