his program would devise some method of generating a reasonably unique fingerprint of a given image and then with a comparison of either new or existing fingerprints be able to identify likely duplicates. These duplicates could then deleted after visual inspection. For reasonably large collection of digital images, my target would be no more than 5% of all duplicates identified would be false matches. After the elimination of the duplicates, I would like for it to create a database of all images, their locations, characteristics and provide a way of classifying the content.
There are several free and commercial programs that do much of the image comparisons. I may end up utilizing one that does this with an Access database and then adding the remaining capability to the database.