Video Synopsis
Background
The amount of captured video is growing with the increased numbers of video cameras, especially the increase of millions of surveillance cameras that operate 24 hours/day. Since video browsing and retrieval is time consuming, most captured video is never watched or examined.Â
Video synopsis is an effective tool for browsing and indexing of such a video. Video synopsis technology can be divided into two categories. The first category is the extraction of key frames, and the second category is video condensation.
Goal
Research and implement the video synopsis algorithms, which contains key frame extraction and video condensation.
Results
Key Frame Extraction
Major trade-off: Reduce redundancy v.s. Keep movement information
K-means method (initialized with Artificial Fish School Algorithm): Minimal redundant frames; but lost timing information and object motion information.
Inter-frame likelihood ratio: Be able to well detect sudden changes in scenes and camera switching. User needs can be adjusted by manually adjusting thresholds.
Motion-volume based frame extraction (novel): Reconcile the two goals (Reduce redundancy v.s. Keep movement information) by weighted summation.
Video Condensation
Definition: It provides a short video representation, while preserving the essential activities of the original video. The activity in the video is condensed into a shorter period by simultaneously showing multiple activities, even when they originally occurred at different times. The synopsis video is also an index of the original video by pointing to the original time of each activity.