Called VisualRank, Google Researchers - Yushi Jing and Shumeet Baluja presented a paper at the International WWW Conference in Beijing detailing on a image search technique that would work on algorithmic detection of components in an image.

An excerpt from the article at TechCrunch

Notably, the new image search technology doesn't just index text associated with an image in determining what's in it. Google is now talking about using computers to analyze the stuff in photos, and using that to associate it in a ranked way with keyword queries. In effect, they're talking about something similar to PageRank for images (but without the linking behavior).

The technique they have used considers each image as a web document with the similarity between images being taken as hyper-links. So the visual links are used to construct a rank for the image and the most relevant will surface when the user performs a query. Moreover, the visual links are computed using different features that are associated with an image - for example, for a people search facial characteristics are used, for landscape color is used.

Image search has its special engines such as Like.com which do product search with a true image analysis. The earlier process for image search was always based on mining the text used to describe images to return relevant results.

Google massive computation power may be the perfect testbed to apply this process and it is to be seen if the problem of automated image recognition is finally near solution.