VisualRank image search was unveiled by Google in Beijing, China today. VisualRank is like PageRank for images in that it uses visual data in addition to text to rank images in the search results pages.

Other companies have been trying to crack the visual search code for years with varying results. Typically, what image search does is use the name of the photo or graphic plus the surrounding text to guess its theme and rank it accordingly.

What VisualRank does is use this data, plus recognize some of the visual data as well. Google is not yet saying which visual data is recognized but one can guess, color, shapes and perhaps even faces and some objects could be identified with VisualRank.

What this means is better results. How many times have you conducted an image search only to find several duplicate images within the top results pages? VisualRank is supposed to identify these matches, but only display the most relevant.

Now, both Google and others are saying that applying VisualRank to its entire database of images would be impractical since the process is processor intensive. But, imagine if VisualRank were applied just to the top 20 results for the most searched for images, how cleaner the user experience would be.

If VisualRank is rolled out on a limited basis such as this, it will greatly enhance image search for visitors, especially those looking for products on the Web. While other image search engines will have visitors seeing double, VisualRank will bring much clarity to an already cluttered field.

Filed under: Google