Super User is a question and answer site for computer enthusiasts and power users. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I'm making a website for a company, only the previous webmaster resized all the images and pasted a watermark on them in the bottom right corner. Now I've got this huge 64GB folder with photos in which I need to find the original versions of the images on the website. Obviously this would take me way too long, and it would be boring work.

Is there anything that can help me speed up this tedious task?

Oh and obviously the file names have changed, too.

share|improve this question
Find one and assume the previous webmaster was systematic? – RedGrittyBrick Aug 8 '12 at 19:33
@RedGrittyBrick The photos come from different locations in that 64GB folder. – RobinJ Aug 8 '12 at 19:35
Any reason you don't use the sort of programs Google shows for "image dedup" - some of them look like they might help for this sort of task. – RedGrittyBrick Aug 8 '12 at 19:38
@RedGrittyBrick… I don't see any programs here. – RobinJ Aug 8 '12 at 19:52
I do – RedGrittyBrick Aug 8 '12 at 20:00
up vote 2 down vote accepted

I have had to do a task very similar to what you describe. The software I ended up using is called Image Comparer. It compares the images "visually" and then lists the ones it thinks match so you can compare them side-by-side (the pic and the file details). They have a free version so you can try it with no obligation. The trial version does just about everything except let you work in batches (so still fully functional, just a bit more tedious to use). I did find that sometimes it would suggest 2 completely different images as being 99% equal, but they are pretty easy to spot so it didn't turn out to be a big deal.

share|improve this answer
Thanks, I'll have a look at it (though I'd first need to set up a virtual machine with Windows, it seems). – RobinJ Aug 9 '12 at 11:41
A lot of what it calls a 98% match is bullshit too, and 94% and lower is almost useless, but this still filters out a lot of them, making the task a lot faster. Thanks! I'm going to leave the question open for a bit in case someone comes with a better suggestion. But for the time being, this will help me a lot! – RobinJ Aug 9 '12 at 13:42

I'd assume you want to write a program to completely automate the process, and the confounding factor will be that even the upper-left corners of the images, which are not polluted by the watermarks, are not identical in the resaved images due to lossy image compression.

Write a program to load and calculate some simple metrics for the upper-left part of each image, which should be nearly identical in the original and watermarked images. Average pixel value, or x,y location of the max or min values. Pair the images by closest match.

share|improve this answer
I haven't done anything with image manipulation yet in any programming language. Also, some of the images have been cropped. – RobinJ Aug 9 '12 at 11:40

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .