Shot detection in video sequences using entropy based metrics

Abstract

ABSTRACT

Video shot segmentation using fusion of SVD and mutual information features

“…Let AE be the angle between Ö and ¼ , then Ö Ì ¼ Ê Ó× (12) where Ê is the resultant length of the resultant vector Ö that corresponds to the -th cluster. The null hypothesis is accepted if [13], [14] Ó× AE ½ ´Ò Ê µ à ½ ´Ò ½µ´Ã ½µ « Ò ½µÊ (13) where à ½ ´Ò ½µ´Ã ½µ « is the upper « percentage point of the F-distribution with degrees of freedom à ½ and´Ò ½µ´Ã ½µ.…”

“…It can be used for detecting abrupt cuts, where the image intensity or color changes abruptly. The mutual information value Á for each two consecutive frames was calculated (as presented in [12]) separately for Ê, and color components. For example, Á Ê for the Ê component is expressed by:…”

“…The first method relies on performing singular value decomposition on a matrix created by the 3D color histograms of single frames [11]. The second method relies on the mutual information between two consecutive frames [12]. Fused (concatenated) features from these two methods are next processed by a clustering method in order to detect the shot boundaries.…”

Audio PCA in a novel multimedia scheme for scene change detection

“…In order to detect the various video shots, the mutual information and the joint entropy between two successive frames was calculated separately on each of the RGB channels, as is proposed in [10]. The mutual information between the frames t f and 1 + t f for the red channel with N gray levels is defined as …”

“…A small value of the mutual information f . In order to detect possible shot cuts, an adaptive thresholding approach was employed, as in [10].…”