We work on the problem of video summarization via a fully automatic structural analysis of the video and audio data. Our solution has three elements:

1. Computing film segments: We solve this problem by analyzing film production techniques as well as the psychology of audition. We also incorporate the idea of a viewer memory model. We use both audio and video data for segmentation. We've also developed algorithms for detecting structural elements such as dialogues. 
2. Stochastic Film Analysis: We've investigated the stochastic characteristics of shots in terms of their duration, descriptive complexities, Analysis of the distributions yield algorithms that enable to sub-sample the original shot sequence. This is done in conjunction with audio analysis and while preserving scene boundaries.

3. Shot Condensation: Given a sub-sampling of shots, how to determine the duration of each shot and the transitions (i.e. fades etc.) between them? This is done by determining an interesting relationship between the descriptive complexity of a shot and decoding time time for that shot.