Video condensation by ribbon carving

Area of Science:

• Computer Vision
• Video Processing
• Digital Forensics

Background:

• Efficient browsing of long video sequences is crucial for applications like visual surveillance and video forensics.
• Existing methods like frame skipping, video summarization, and video montage have limitations in preserving temporal or semantic context.
• Video synopsis offers improvements but is complex and involves multiple processing stages.

Purpose of the Study:

• To introduce a novel and conceptually simple method for video condensation.
• To develop a technique that efficiently browses long video sequences while preserving context.
• To offer an alternative to complex video summarization and synopsis methods.

Main Methods:

• Introduction of the "video ribbon" concept, inspired by image resizing techniques.
• Recursive carving of ribbons by minimizing an activity-aware cost function using dynamic programming.
• Development of a flexible ribbon model allowing adjustment of condensation ratio and event anachronism.
• Proposal of sliding-window ribbon carving for streaming video.

Main Results:

• Video condensation is conceptually simple and relatively easy to implement.
• The method effectively removes information from the space-time video volume.
• The ribbon model allows flexible trade-offs between temporal condensation and event anachronism.
• Demonstrated efficiency on motor and pedestrian traffic data.

Conclusions:

• Video condensation presents a novel and efficient approach to browsing long videos.
• The proposed method overcomes limitations of existing techniques by preserving context.
• The technique is adaptable for both offline and streaming video analysis.