Image Search Redefined
How to hunt down interestingness by its hexadecimal color.
By Maria Popova
It’s been a while since we’ve stumbled across something along the lines of PicLens, retrievr and the Flickr Related Tag Browser.
Today, we bring you another inspired algorithm that revolutionizes the image search experience.
The Multicolr Search Lab, offers color-based search, spitting out images in up to 10 colors you’ve specified. Out of the equally inspired Idée Labs, a self-proclaimed “technology playground for visual search,” Multicolr utilizes the proprietary Piximilar visual similarity search technology that scours large collections of images without using keywords or metadata.
The Multicolr Search Lab is currently available for Flickr and Alamy Stock Photography. The Flickr version extracts the colors from over 10 million of Flickr’s most interesting Creative Commons images. The notion of “interesting” is actually one of Flickr’s own cool algorithms that assesses an image’s “interestingness” based on various meta elements like where the clickthroughs are coming from, who comments on it and when, who marks it as a favorite, its tags and more.
Because these factors are in constant flux, so is the “interestingness” of any given image — something meant to inspire more exploration and discovery inside Flickr.
And while our 8th-grade English Lit teacher used to say that “interesting is what you call an ugly baby,” we have to admit this brand of interestingness falls squarely on the baby pagent side.
Published October 6, 2008