Major Findings

1. Steganography: we have shown that, over a broad range of natural images, certain higher-order statistics in a wavelet decomposition are highly similar. We have also shown that when a message is embedded into an image, in a manner that is imperceptible to the human eye, these statistics are dramatically altered. As such, we are able to detect the presence of covert communication.

2. Digital Tampering: we have developed several techniques for determining if an image has been tampered with from the time of its recording. These approaches work by observing that certain forms of tampering leave behind specific statistical correlations that can be quantified and detected.

3. Computer Graphics or Photographic: Computer graphics rendering software is capable of generating highly photorealistic images that are often very difficult to differentiate from photographic images. We have, however, developed a method for differentiating between photographic and computer generated (photorealistic) images. Specifically, we have shown that a statistical model based on first- and higher-order wavelet statistics reveals subtle but significant differences between photographic and photorealistic images. This work has begun to have interesting legal applications in light of the 2002 United States Supreme Court ruling that effectively legalized "virtual child pornography".

4. Art Forgeries: We have developed a mathematical technique that can classify various parts of a painting as belonging to one or more artists. Beginning with a high-resolution digital scan, our technique works by looking for statistical differences across the canvas. We have also applied this technique to detecting art forgeries - in collaboration with the Metropolitan Museum of Art, we have analyzed drawings by Bruegel and find that we are able to perfectly distinguish between authentic drawings and known forgeries.

5. Virtual Reality and Architecture: we have previously digitally reconstructed an ancient Egyptian tomb from a series of photographs. We have applied these techniques to digitally reconstruct the Orozco murals at Dartmouth, Pomona College and New School University. This work was part of a larger Orozco exhibit at the San Diego Museum, allowing visitors to experience, in 3-D, the beauty of these powerful murals. This exhibit was also at Dartmouth's Hood Museum.

6. Medical Imaging: we have developed an elegant and powerful computational framework for the registration of medical images. This technique overcomes many of the limitations of existing approaches.

7. Neural Binding: we have provided experimental evidence that suggests that a particular model of neural binding, based on temporal synchrony, has a serious flaw. This work is particularly important as expensive and time consuming experiments may be devoted to the search of neural mechanisms, that we believe simply don't exist.