We have talked about reverse image searching with TinEye here in the past – here is some recent research that takes the concept further. John Resig collaborated with the Frick Art Reference Library to analyze images lacking identification from their collection. Using software developed off the TinEye model he was able to establish trends, and ultimately relationships, with identified images elsewhere. This foray into ‘computer vision techniques’ proved to have a higher rate of success than anticipated, resulting in a large number of identifications of images previously lacking information on creator or location.
Read Resig’s detailed and fascinating report here, and also read about his Ukiyo-e Search project here.