Video key frame selection by clustering wavelet coefficients

This paper presents a new method for selecting key frames from a given video sequence. It is characterized by the fact that the entire process works in a wavelet transform domain. At first, shot boundaries are detected to define initial key frames. Secondly, specified but any number of key frames are selected by clustering feature vectors. Its effectiveness is evaluated in terms of precision rates and processing speeds. The proposed method offers more satisfactory results and works faster than the other existing methods.