Quote:
Originally Posted by Nautilus
This digital fingerprinting tech is covering all stuff including those videos already circulating. It's kinda comparing two videos pixel by pixel to see if they match, just more sophisticated. You do not need to add any watermarks (visible or invisible) to your video to make it identifiable by this tech - it works with any video as is. And since it's pixel-by-pixel comparing, it can find matches even if the original video was strongly modified, for example recompressed, cropped, rewatermarked etc.
This thing is cool, it's just very server intensive - that's why they're not offering the whole lot of tubes with their monitoring services, and limit the amount of videos that you can submit for a fixed price. This thing is just too heavy on CPU usage to let it loose scanning all infringing sites comparing all finds with the database of a million protected vids and charge $10/month for that.
Aside from CPU issues, the cost of a license is still out of this world for most DFP apps - one more reason why all DFP services charge insane prices.
http://en.wikipedia.org/wiki/Digital...fingerprinting |
Well, it isn't actually pixel by pixel - it's actually based on binary hashing originally, to circumvent compression technologies which is stripping down an images components by converting RGB to YCbCr format for hash extraction eg
All derivations of fingerprint technology are based on this. The new stuff invovling videos is grouped hashing of multiple frames, but I don't believe any are based on hashing of the entire video, just a few hundred frames...
It's not really very CPU intensive though - what is (and bandwidth intensive too) is detecting it, once the video has "gone out"