Embedding Imperceptible Data in Images

Research Overview

By altering the colours of an image by imperceptible amounts, one can embed digital information that is easily detected by common desktop scanners. But people cannot see the image distortions. My current work uses quantitative models of human visual discrimination to constrain the visibility of embedded information, along with conventional image processing techniques for its detection. This is a collaboration with Todd Cass and David Hecht at Xerox PARC, and with David Heeger (Stanford University).

Applications of this technology are many, from the embedding of calibration information in images (e.g. for deskewing), to digital watermarking, to the embedding of web addresses for example.

Related Publications

Return to David Fleet's home page.