An algorithm of key-frame extraction based on adaptive threshold detection of multi-features

Key frame extraction is an important component of content-based video retrieval and directly influences on the efficiency of video retrieval. Nowadays some problems are existed in the algorithms of key-frame extraction, such as features selected singly, choosing threshold value difficultly and so on. This paper proposes a new key-frame extraction method which based on adaptive threshold detection of multi-features. First, two descriptors of color histogram, edge histogram of the adjacent scale wavelet transform are used to describe visual content, and combine to form a frame difference measure of multi-feature integration. Then key-frames are extracted by the adaptive threshold detection. Finally, experimental results show that the proposed algorithm is efficient.