At the International World Wide Web Conference, held last week in Beijing, China, Google’s researchers presented a paper on PageRank for Google Images illustrating Google’s attempts to improve search results for photos, art and graphics. The new Visual Rank algorithm combines image-recognition software methods with techniques for ranking images that look most similar.

Image search is a popular feature in many search engines but the most of image searches use insufficient image information to rank the images. For the most part, results are merely generated through the text clues of the pages in which the image is embedded and often ignore the content of the images.

Globally, computer scientists have made their best but the problem with image analysis still persists. Google’s VisualRank does not contribute in this direction. The desired functionality for search engines to identify people or “read” activities in a photo is still unattainable.

Here is how Google’s researchers explained their approach: “Through an iterative procedure based on the PageRank computation, a numerical weight is assigned to each image; this measures its relative importance to the other images being considered. The incorporation of visual signals in this process differs from the majority of large- scale commercial-search engines in use today.”

For the purposes of its paper, entitled “Page Rank for Product Image Search”, Google conducted a series of experiments by retrieving images for 2,000 of the most popular products queries in Google. Users in the experiments felt more satisfied by the results and their relevancy.

The considerable benefits of Google’s new approach will comprise the reduction of image duplicates in search results as well as the significant decrease of the irrelevant images displaying for the particular query.

Since Google’s Universal Search is gaining popularity, the new VisualRank will play its important role in SERPs. Definitely, a visual ranking system will improve image search quality and bring additional utility to the users because more relevant images will result in better search for everyone.

To get more on Google’s new algorithm by following the link: http://www.esprockets.com/papers/
www2008-jing-baluja.pdf

Copyright © 2010 DevStart, Inc. Permission is required to use the material on this page.