Detection of abrupt and gradual change of scene in a video is an active area of research for last many years due to its applicability in many applications, viz. content based image retrieval, computer vision, etc. This paper presents dynamic dual threshold based schemes to detect the abrupt scene change in a video. Proposed schemes use static as well as dynamic thresholds defined over set of features proposed in this paper. These features include Mean Squared Error (MSE), Entropy, and Count of displaced blocks. Extensive experiments have been conducted to recommend the values for the proposed thresholds to efficiently detect the abrupt scene change in videos. Experimental results show the good accuracy to detect the abrupt change of scene.
展开▼
