Apr 22 2009

Google Labs Release Similar Images Search

Google Labs is an infrequently updated part of Google that showcases new or interesting things that Google are working on. One thing that was of interest to me recently was the Similar Images search feature. This will allow you to search for a term using the normal Google image search, but adds the option to click on a link next an image and view more images that look like this one. For example, lets say you wanted to search for images of London, you can click on images of the London Eye and see different images of the same thing. Here is the official video from Google.

I thought I would have a play with this feature and see what I could do with it. One thing I always have trouble finding is an image of a mouse cursor. When I am writing user manuals I like to have the cursor in the image so that the user can see where they are mean to click. So after a image search for cursor I found the following results.

Google Image Search For Cursor

Google Image Search For Cursor

I then clicked on the similar link for the 6th image along on the second row and got the following page.

Google Image Search Similar Images

Google Image Search Similar Images

As you can see the feature if nearly there. From this page of images quite a few are indeed cursors, but the majority are not at all like the original image. For some reason Google seems to like to display images of cupid for this particular search.

I had more success when I tried searching for logos, which I presume if because they look quite alike. Searching for the logos of Wordpress, BBC, and Microsoft gave some good results.

So the big question is Will this become part of the main image search? I think the answer is probably yes. There are a quite a few products and features developed over the years that have become live, such as Gmail and iGoogle, so I think that we could very well see this feature appearing on normal Google image search results. You can already view images by face and drawing type, so this is probably just another extension of this. I just hope they refine the image recognition before putting it live.

Philip Norton
Lead Developer, Research and Development