Recommendations are independently chosen by Reviewed’s editors. Purchases you make through our links may earn us a commission.
*August 2, 2006 – *A Massachusetts Institute of Technology and University of Toronto research team debuted photo-deblurring technology at yesterday’s 33rd Annual Siggraph Conference in Boston, MA.
In a seminar entitled "Removing Camera Shake from a Single Photograph," the MIT – U of Toronto research team presented an algorithm to correct high-level blurs at the world’s largest electronic and computer graphics conference today, which hosts "the best and most senior minds in technological innovation," according to Siggraph spokesperson Brian Ban.
The anti-blurring mathematical model can only correct a particular type of blur caused by hand motion. As digital cameras and camera cell phones gradually shrink in size, the model can accommodate the recent phenomena of hand motion with lightweight cameras.
The algorithm is based on the principal that slight hand motions of even only a few millimeters cause camera rotations, resulting in image blur according to researcher and post-doc Rob Fergus in the MIT Computer Science and Artificial Intelligence Lab.
The camera shake is modeled as a blur kernel through a technique called deconvolution, according to the abstract, "Removing Camera Shake from a Single Photograph." Unlike other image software that simplifies blur kernel formation, the new technique infers blur kernels by estimating the distribution of a number of probable images, according to the abstract. Unlike the Adobe Photoshop unsharp mask tool which simply adds contrasts to the image, the algorithm can correct and sharpen the image, said Fergus.
Geared for users of small handheld cameras, the new post-production technique could eliminate the need for bulky tripods or help camera owners who lack popular (and costly) anti-shake features in newer point-and-shoot cameras. "With lighter, newer cameras, [blurry images] are a common thing. We normally delete them because we don’t know what to do with them," said Fergus, but the "photos you really care about" can be saved.
The mathematical model, however, cannot correct other kinds of blur, including inadequate depth of field in which images are out of focus. The model also cannot compensate for slow shutter speeds for fast moving objects such as cars.
Although this technology is "not the final solution to all types of blurs," according to Fergus, "There are enough situations where it can be used."
"This is a research paper," Fergus notes, "we’re not going to see this next week in Photoshop," although the MIT post-doc states the algorithm would be ideally applied to the Adobe software program within another year or two.
The final research paper was completed in May 2006. At the time of interview, Fergus said patent had not yet been filed.