Sign In

Communications of the ACM

ACM TechNews

Deep Learning Can Now Flawlessly Correct Photos Taken in Almost Complete Darkness

View as: Print Mobile App Share:
An underexposed image corrected using deep learning algorithms.

Researchers at Intel and the University of Illinois UrbanaChampaign have developed a way to use machine learning to artificially boost the brightness of a dark photo.

Credit: University of Illinois UrbanaChampaign, Intel

Machine learning can artificially increase the brightness of photos taken in low-light conditions with impressive results, according to researchers from Intel and the University of Illinois, Urbana–Champaign.

Using deep learning techniques, the team trained an algorithm to properly brighten and color-correct a poorly exposed image during post-processing.

With the new algorithm running on a smartphone's camera app, the small sensors inside smartphones may not need to match the performance of the sensors used in digital single-lens reflex (DSLR) cameras. Instead, a software update one day could allow smartphone photos from the average consumer to rival images from the most expensive cameras.

From Gizmodo
View Full Article


Abstracts Copyright © 2018 Information Inc., Bethesda, Maryland, USA


No entries found