Video Montage

Unedited video often has significant content that is of low interest to the viewer. One way to present such videos is by creating a video montage preserving interesting content. Manually doing this is difficult for a non-expert without specialized video editing software. We propose and demonstrate a technique for creating a video montage automatically from unedited video with minimal need for user intervention. Interestingness in a video is inherently subjective, therefore to incorporate the user’s notion of interestingness we ask for user input. This constrains us to using a small amount of data for training. The scenario that we have in mind is: the user browses through a few segments of the original video, marking them as interesting or uninteresting. We use this labeled data to train a classifier about what is interesting to the user, and then classify the rest of the video. The video classified into interesting and uninteresting segments is then processed to join the interesting segments and to shorten the uninteresting segments.