Searching for images on the web could be improved thanks to a technique developed by Google.
Picture searches currently rely on text cues to help decipher the image but, according to Google researcher Shumeet Baluja, the majority of image searches hardly use any image information to rank the pictures. Instead search engines such as MSN, Yahoo and Google itself rely on text on the pages in which the image is embedded. This can, however, throw up unrelated results.
With this in mind, Google researchers developed VisualRank, an algorithm that identifies themes in pictures and then ranks images on how similar they look to other photos that contain the same themes.
Simply counting the number of similar visual features isn't enough, however. According to Baluja a mechanism that utilises the information for the purposes of ranking is also required.
"Simply counting the number of common visual features will yield poor results. To address this task, we infer a graph between the images, where images are linked to each other based on their similarity. Once a graph is created, we demonstrate how iterative procedures similar to those used in PageRank can be employed to effectively create a ranking of images," said Baluja in a research paper.
"We implicitly rely on the intelligence of crowds: the image similarity graph is generated based on the common features between images. Those images that capture the common themes from many of the other images are those that will have higher rank."
In the paper, Baluja claims that when 2,000 Google employees compared VisualRank results with those from a standard image search, the new algorithm returned 83 percent fewer irrelevant images.