Sunday, April 27, 2008

Image Search – Goodbye Alt Tags!

Currently Google image Search uses the standard SEO-familiar methods of ranking all images – alt tags, file name etc. There is also an element of human analysis with a random image review system. This is how it has worked for quite some time, but perhaps not for much longer.


Last week at the World Wide Web Consortium conference in Beijing*, two Google scientists delivered a paper which could revolutionise image search and turn at least one element of Universal Search on it’s head.


The paper, “Page Rank for Product Image Search”, outlines a method to remove Google’s dependency on alt tags for image description. The paper proposes a new system called ‘VisualRank’. This new algorithm takes elements of cutting edge image recognition software and incorporates existing techniques for weighting and ranking. This new type of image search will be able to rank images directly from the image contents.


While the algorithm is still in testing and not yet ready for commercial use, it may not be too long before there is no human interface needed to create extensive image ranking. Crawlers, using the new system, will work across the web devouring every un-optimised image out there. Eventually you should be able to find a picture of literally anything!



*It will probably seem very surprising to hear that China (Beijing) will be hosting a conference on the web's future, themed 'One World, One Web'. What is probably even more surprising is that the World Wide Web Consortium (W3C), which oversees web standards, is lending its credibility to the conference.

No comments: