Both Google and Hitachi America now offers services to search for images using image recognition to deliver better results. But Hitachi's services are no longer notch in this area than what Google is now doing.
Part of the challenge of image search on the web is that there are many images out there that are not associated with any form of relevant, descriptive information. Thus, a search service does not necessarily know what the image contains. Thus, the search results may be missing a lot of pictures that really are relevant.
In other cases, the keywords that the user type, are ambiguous. Printer-user in the word "jaguar", it is unclear whether the author believes, for example, cat or car.

Google is now starting to roll out a feature that allows you to find images similar to the one you already have gotten out of the search results. This is called "Similar Images" and has until now been an experimental service under the Google Labs.
Google Image Search with Simula Images
Google Image Search with Simula Images
It works so that you can click on a special link that appears under each of the images in search results. Then you will come to a results page with images that are similar.
A closed beta version has been in operation since last September.
On GazoPa given user to search for images based on keywords, then to display images similar to the one you choose from the results. Hitachi is behind the service GazoPa, which yesterday was made available in a public beta.

It can, however, discussed how good the results are. When we uploaded a picture of an upcoming mobile phone, we got up in the search a lot of pictures of similar phones, but also some pictures of model cars with about the same color as your mobile phone - quite independent of the settings we chose for the search. So you can say that it remains a bit before GazoPa is ready for official launch.