An Automatic Cut Detection Algorithm Using Median Filter And Neural Network
Seung-Chul Jun, Sung-Han Park · 2002
In this paper, an efficient cut detection algorithm in MPEG encoded bitstream is proposed. The proposed method distinctly separates between cuts and not-cuts by applying median filter to the histogram difference and pixel difference representing the variation of two successive frames. Then cut and not-cut are classified using the k-means clustering algorithm without a threshold value. This classification provides the input training set of the neural network. Since the measurements retrieved in this way are distinctly separated between cut frames and not-cut frames, the cut frame can be detected using back-propagation neural network without miss-detection and false positive. 1