Copy detection for digital media is critical to preventing copyright violations and enforcing copyright. Watermarking is the most widely accepted form of detection. However, there are two significant limitations of the watermarking approach. First, Since watermarks must be introduced into the original content before copies are made, it cannot be applied to content which is already in circulation (legacy materials). For example, if we wanted to find all clips of Star Wars posted on the web, watermarking would not provide a solution. Second, in the eventuality that the watermark on a particular piece of content is compromised, there is no alternative approach to copyright enforcement.
Content Based Copy Detection (CBCD) is a complementary technology to watermarking which provides a solution to the two problems mentioned above. The primary thesis of CBCD is the media is the watermark. That is, the media (image, video, audio) contains enough unique information to be able to detect copies. There are a number of commercial applications of this technology:
Our CBCD system works by first breaking the commercial video segment into a number
of characteristic keyframes as shown above. It then
computes a simple edge and color based measure on each frame and
stores the resulting sequence of measures as a signature for the clip in an index table.
When live video is played into the system, it performs the same keyframing and signature
generation on the incoming material and looks for likely matches to its stored material.
The implemented CBCD system can already match a live video feed (in real time) against a
collection of approximately 2 hrs of video on a 500 Mhz Windows NT machine. Currently
tests are in progress with larger collections on the order of 10 hours.
VideoGREP: Video Copy Detection using Inverted File Indices
Arun Hampapur, Rudolf M. Bolle
Technical Report on some unpublished work on index compression for copy detection
More
| Contact: Arun Hampapur | Last updated: 6/12/02 | ||
 |
 |
 |
 |