‘Reversible data hiding’ embeds data in pictures

Scientists from the University of Rochester and Xerox have invented a new way to hide information within an ordinary digital image and to extract it again – without distorting the original or losing any information.

Scientists from the University of Rochester and Xerox Corporation have invented a new way to hide information within an ordinary digital image and to extract it again – without distorting the original or losing any information.

Called ‘reversible data hiding,’ the new technique will solve a dilemma faced by digital image users, particularly in sensitive military, legal, and medical applications. Until now they have had to choose between an image that’s been watermarked to establish its trustworthiness and one that isn’t watermarked but preserves all the original information, allowing it to be enlarged or enhanced to show detail. When information is embedded using the newly discovered method, authorised users can do both.

The technique, described in a paper presented at the IEEE 2002 International Conference on Image Processing in Rochester, NY on September 24, was co-developed by Mehmet U. Celik and A. Murat Tekalp of the University and Gaurav Sharma and Eli Saber of Xerox. Their collaborative research was done in the Center for Electronic Imaging Systems (CEIS), a New York State Office of Science, Technology and Academic Research designated centre for advanced technology.

‘Commonly used techniques for embedding messages such as digital watermarking irreversibly change the image, resulting in distortions or information loss. While these distortions are often imperceptible or tolerable in normal applications, if the image is enlarged, enhanced, or processed using a computer, the information loss can be unacceptable,’ said Gaurav Sharma, an imaging scientist at Xerox’s Solutions and Services Technology Center in Webster, N.Y.

‘With our new data embedding algorithm, authorised recipients not only can extract the embedded message but also can recover the original image intact, identical bit for bit to the image before the data was added,’ he said. ‘The technique offers a significantly higher capacity for embedding data and/or a lower-distortion than any of the alternatives.’

‘The technique will be widely applicable to situations requiring authentication of images with detection of changes, and it can also be used to encode information about the image itself, such as who took the picture, when or with what camera,’ said Murat Tekalp, professor of electrical engineering at the University.

‘The greatest benefit of this technology is in determining if anyone has clandestinely altered an image. These days many commercial software systems can be used to manipulate digital images. By encoding data in this way we can be sure the image has not been tampered with, and then remove the data within it without harming the quality of the picture,’ he said.

Although the technique is currently implemented in software, it could be implemented in hardware or firmware in trusted devices where image integrity is critical to the application, the authors said. For instance, the technique could be used in a trusted digital camera used to gather forensic evidence to be later used at a trial. If information is embedded in the images captured with the camera using the new algorithms, any subsequent manipulations of the pictures could be detected and the area where they occurred pinpointed.

A patent application on the methods developed for reversible data hiding has been filed by the University of Rochester; and the University and Xerox will share the rights to the invention.