Tsallis entropy-based information measures for shot boundary detection and keyframe selection

摘要: Automatic shot boundary detection and keyframe selection constitute major goals in video processing. We propose two different information-theoretic approaches to detect the abrupt boundaries of a sequence. These are, respectively, based on information measures, Tsallis mutual Jensen–Tsallis divergence, that are used quantify similarity between frames. Both measures also find out most representative each shot. The representativeness frame is basically given by its average with respect other frames Several experiments analyze behavior proposed for color spaces (RGB, HSV, Lab), regular binnings, entropic indices. In particular, HSV Lab only 8 bins component an index 1.5 1.8 substantially improve performance previously methods Jensen–Shannon divergence.