Abrupt Shot Boundary Detection from Video Sequence Using Motion Direction Histogram Feature

We propose the motion direction histogram (MDH) as a new feature for abrupt shot boundary detection in video sequences. Conventional methods for abrupt shot boundary detection use a threshold difference as a measure of the feature values between consecutive images. It is difficult for such conventional methods t o detect shot boundaries in "busy" scenes, in which intensities change substantially from frame to frame, because these methods use features such as color or luminance difference histograms and motion vectors. The values of these kinds of features are significantly increased when a video sequence includes such intense motions. To overcome this problem, we have developed the MDH feature. We confirmed its effectiveness experimentally and obtained an accuracy of over 95 % for abrupt shot boundary detection.