Image Databases to Find “Visually Similar” Imagery and More

Some image search news and tools that you might have missed.

Image search provider, Cydral (based in Paris, France), has released a new version of their image database (they claim about 150 million images). What makes the Cydral service different (as far as imagery from the open-web goes) is that it allows the searcher to find similar images based on “visual similarity” (color, texture, shapes, objects, etc.). Here’s an example. A search for “waterfall” and then a click on the “similar images” link for the third image listed. The technology is far from perfect but it’s an example of a service that many web image engines will be providing in the future.

Want more info and demos?

In January I posted about LTU Tecnologies, one of many companies doing work to allow the user to find similar imagery based on the composition of an image. You’ll also find a link to LTU’s Image-Seek This demo database allows you to search a database of about 65,000 royalty-free images from the Corbis catalog and then proceed to find visually “similar” imagery.

Another interesting and somewhat related image search service (it’s been available for several years) comes from Cobion via the German site, (translated into English via Yahoo). The technology here allows you to search for text found inside an image. For example, if the word “London” is found in an image of a map, this tool might allow you to find and view the map. Also, you can limit your search to images that contain the face of a human.

Postscript: VIMA Technologies is another company that licenses search technology to find visually related imagery. You can demo here and here.

Related reading