Detecting duplicate videos

[MrCrispy]

I turn to the experts and hope something like this exists. It should detect duplicates by analyzing frames in the video and be able to identify clips that share parts of the video (e.g. a movie rip, and  youtube clips from the movie, ) and mark the ones with the highest resolution and quality.

I found research papers that propose methods of doing this -

http://scenic.princeton.edu/ppt/edge-lab-video.pdf
http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=5583860&userType=inst

Obviously this will be very expensive computationally but it doesn't matter.

[4wd]

I suppose as a "cheap" way of doing it, you could:
1) Convert both clips into individual frame images.
2) Compare using an image comparison program, ideally one that would just output a report - if all the images in the shorter clip compare at, say, an 80+% match with some of the images in the other - chances are it's a clip cut from the other, (either it or its source).

Downside is it will be hugely resource intensive in terms of space, time and CPU/GPU.

Here's a script on Ubuntu, (and a link to a Ruby version in the last post), that will compare exact same video across different formats/framerate: Script to find duplicate videos based on content regardless of format. - it might give you some ideas.

There's also UQLIPS, (video upload/search link): A real-time near-duplicate video clip detection system (PDF paper) - who knows, they may be working on a standalone program.