Communications of the ACM
New Algorithm Repairs Corrupted Digital Images in One Step
Researchers at the universities of Maryland and Bern have developed an algorithm that uses artificial neural networks to apply a wide range of fixes to corrupted digital images.
Credit: Don Cochran, Kodak Lossless True Color Image Suite
Researchers at the University of Maryland (UMD) and the University of Bern in Switzerland have developed an algorithm that incorporates artificial neural networks to simultaneously apply a wide range of fixes to corrupted digital images.
The algorithm can be trained to recognize what an ideal, uncorrupted image should look like, enabling it to address multiple flaws in a single image.
The researchers tested the algorithm by taking high-quality, uncorrupted images, purposely introducing severe flaws, and using the algorithm to repair the damage. In many cases, the new algorithm outperformed other conventional methods, very nearly returning the images to their original state.
Other researchers have used artificial neural networks to address problems piecemeal, but the new algorithm can simultaneously address a wide variety of problems, notes UMD professor Matthias Zwicker.
The researchers trained the algorithm by exposing it to a large database of high-quality, uncorrupted images widely used for research with artificial neural networks.
From University of Maryland
View Full Article
Abstracts Copyright © 2017 Information Inc., Bethesda, Maryland, USA
No entries found