Identification of editing effect in image sequences by statistical modeling

This paper presents a novel approach for video editing effect detection in digital image se- quences such as movies, commercials... The proposed method is based on histogram information and the decision criterion is derived from statistical properties of dissolve or wipe images, with respect to some hypothesis on the histograms of shots that surround the dissolve interval. More in particular the histogram of a dissolve image is given by the convolution of the histograms of two contiguous shots, provided that the random processes associated with the two shots are sta- tistically independent. A two step algorithm is presented: in the first step the dissolve effects are detected while in the second a precise estimate of the editing effect time interval is established. Simulation results indicate that high performance can be achieved with the proposed approach.