Question about the possibility of a reverse image lookup

Post by **tnf** » Tue Jul 26, 2005 6:04 pm

Do any of you programmer types know if it would ever be possible to develop some sort of image lookup - where you could basically check the entire web to see where the orginal source of an image being hosted somewhere else is?

For example - let's say a pic is on imageshack, and you want to find its original source or something (assuming it was taken from another website).

The only way I could see this being done is to run check the actual information contained within the image file and then compare it against the entire web - something that would obviously be a real challenge at the moment.

But do any of you know if any technology like this is being considered in light of the amount of copyrighted material floating around out there that is not supposed to be?

Post by **Foo** » Tue Jul 26, 2005 7:31 pm

There are image comparison algorithms which are very sound. Thumbs plus has one, google evidently have one.

The question is whether anyone could release it mainstream. I mean, imagine the implications of being able to input someone's face and find the yearbook it came from.

Post by **Survivor** » Tue Jul 26, 2005 8:25 pm

Don't code but if you would take some average lines like top, middle, and bottom line of an image which would be smaller than the actual image but still contain a reasonable precise amount of data to compare against.

Post by **duffman91** » Tue Jul 26, 2005 9:14 pm

The only way to deal with that is through watermarks.

However, they can be removed.

You're asking for something very doable, but not enforsable or convenient.

Post by **duffman91** » Tue Jul 26, 2005 9:15 pm

Foo wrote:There are image comparison algorithms which are very sound. Thumbs plus has one, google evidently have one.

The question is whether anyone could release it mainstream. I mean, imagine the implications of being able to input someone's face and find the yearbook it came from.

Image comparison and a database of image information are two different things.

Post by **Foo** » Tue Jul 26, 2005 9:42 pm

Image watermarking is not the only way to deal with this.

As an example earlier, take a look at the thumbsplus algorithm for image comparison for an indication of this. It'll work on images of varying size, adjustments (colouration, saturation etc) and often even crops/warps.

Why did I cite google? Not for the cataloging, but for the suite of applications they've got. Google images and picasa will be the tip of the iceberg as far as their imaging sophistication stretches. IIRC, doesn't picasa have an image comparison tool similar to thumbsplus' system?

..and the systems I'm listing here are maybe 6 years or more old.