magicfern
magicfern copied to clipboard
Question: can fingerprints be compared to each other, or do you always need a reference image?
Hello, it is possible (likely) that I'm misunderstanding how this library works; i understand that:
- to create a camera fingerprint, you should use 20+ images from it
- then you can use that fingerprint to determine how likely it was that the camera took any given photo
But is there a way, if you've built up a database of fingerprints (not sure how this works), to look up other fingerprints by similarity, essentially to see if this camera's fingerprint has already been added to the database?
Hey, sadly you cannot do this directly. As implemented in candidate_processing.cpp in line 135 and 139, you first need to calculate the cross correlation and then the PCE. Calling getCrossCorrelation you need to make sure to provide the fingerprints in the correct format.