SnapperTalk

When photographers need to search for their photos on the internet there’s a few ways available, such as Google Images and other search engines. What they all share in common is these methods search by the text information associated with the image – whether it be the photo caption itself or the content of the page the image is on.

Fellow photographer Jason Bye tipped me off to a new image search engine called TinEye, currently in private beta, that adopts a radically new method that does not consider the text information at all. Instead it analyses the actual photo itself – “TinEye instantly analyzes your query image to create a compact digital signature or ‘fingerprint’ for it. TinEye searches for your image on the web by comparing its fingerprint to the fingerprint of every single other image in the TinEye search index.”

You can use the search engine in two ways – either by uploading an image stored locally on your computer, or by cutting/pasting a link to an image posted on the internet, and then asking it to search for all other instances of that image.

How could that work? Surely after small differences such as cropping, toning, and other image adjustments, it wouldn’t be recognised as the same image? I was quite skeptical about this point but it does work, and rather well. The search software is capable of identifying images even after they have been heavily cropped, toned, converted to black & white, even if they have text overlaid on the photo e.g. on a magazine cover.

The search engine is currently in private beta which means you have to sign-up to request a beta invite, however I received mine within a day or so. The image-recognition software behind the search engine is also used in the company’s commercial product PixID, which is currently being used by a number of large photo-agencies and other media organisations to track unlicensed use of images on the web. They even have a nice plugin for Firefox or Internet Explorer to enable you to rapidly initiate a search of images.

Those concerned about the copyright of their own images uploaded when using Tineye should note that the terms of service explicitly state that “Copyright for all images submitted to TinEye remains with the original owner/author, and images submitted to TinEye for searching are not added to our index.”

The main limitation I would say is the limited number of images currently indexed – currently about 701,666,310 – sounds like a lot but still misses much. Presumably as the service continues that will improve.