The ImageRover system combines textual and visual statistics in a single index for content-based search of a WWW image database. Textual statistics are captured in vector form using latent semantic indexing (LSI) based on text in the containing HTML document. Visual statistics are captured in vector form using color and orientation histograms. To initiate a search of the ImageRover index, the user specifies a few keywords describing the desired images. The user can then refine this initial query through relevance feedback. During relevance feedback, both visual and textual cues are combined to gain better search performance.
Last Modified: April 20, 1998