The New York Times is reporting on a new research paper about Google’s new image ranking algo which apparently associates an inferred linking relationship between images and uses the PageRank method of iterating ranking values across the graph to come up with final ranking values. This “VisualRank” method was presented in a paper at the International World Wide Web Conference in Beijing this past Thursday, and the process was also reported  at Techcrunch.
Google’s advancements in Image Search
could help keep high-value image results
like this coastal pic stay high in the SERPs
for apropos keywords, while making less-
important images rank far lower.
The new methodology is apparently very adept at weeding out less-important and less-useful images from the search results.
I have earlier reported on Google’s research into Supervised Multiclass Labeling  (“SML”) which can assist with associating keywords with the actual content found within digital images. See also Search Engine Land’s article on Google’s VisualRank Paper .