Method and device for automatic generation of summary of a plurality of images

Abstract

Method and device for providing a summary of a plurality of images, e.g. a video sequence. The method includes dividing the video sequence into a plurality of segments. The segments are analyzed with respect to content and a set of content descriptors are associated to the segments. Preferably, additional textual information about the segments, screenplay etc., is used to determine the content descriptors. A graph representing relations between the segments is constructed indicating relations between segments. Weights are associated to the relations so as to represent a measure of relation, e.g. a logical correlation, between segments. The weights are based on the calculated content descriptors. A relevance measure for a segment is determined based on all weights associated with relations to said segment. Finally, a summary is generated by selecting the most relevant segments. The method can create an automatic summary of a film that preserves all the logical plot of the original but is shorter in duration (e.g. 70% of the original film) while the original playback rate is preserved.