TinEye isn’t really an image search engine — it’s not particularly useful if you’re looking for pictures of some a subject like puffins or the Brooklyn Bridge, or if you’re looking for pictures of a particular person, like Millard Fillmore. TinEye is designed for a totally different purpose. You show it a picture, it analyzes it and then looks in its index for other copies of the same picture. It’s supposed to be able to find other versions, even if they have been cropped or altered.
How does it work? According to the TinEye FAQ:
Using sophisticated pattern recognition algorithms, TinEye creates a unique and compact digital signature or ‘fingerprint’ for each image that is added to the index. When you submit a search image to TinEye, its fingerprint is analyzed on-the-fly and compared to the fingerprint of every single image in the TinEye search index.
In other words, it’s magic.
This could be a very useful tool for finding unauthorized copies of photographs and other images, and I’ve seen people on the Flickr forums who have had success with this. Unfortunately, the database is just too small at the moment to be really successful with this, at least based on my testing. I searched about twenty of my most Flickr photographs, including several that I know have been used on other sites, and the Big Blue Bear photograph is the only match I’ve found. But this definitely has possibilities for copyright issues and other situations where you want to track how an image has been used.
They have also used it for some cool searches showing altered versions of famous paintings and other iconic works — Hello Kitty, the Linux penguin, the Abbey Road album cover, etc. Here’s a widget showing different versions of the Mona Lisa: