NVIDIA AI scrubs noise and watermarks from digital images

Brittany A. Roston - Jul 10, 2018
NVIDIA AI scrubs noise and watermarks from digital images

NVIDIA researchers are back with yet another digital image technology that pushes the limits of traditional image manipulation. Unlike Adobe’s recently disclosed project, which involved a neural network trained to spot digitally altered images, NVIDIA’s newest creation can scrub digital sensor noise and watermarks from digital images. Thanks to the artificial intelligence powering it, the feature is far more effective than existing denoising tools.

Digital camera sensor noise, though not as severe as it once was, is still common from consumer-tier cameras, particularly smartphone cameras in low-light conditions. This is due to the small sensor size used in these cameras, making post-processing necessary to increase image quality. Some products, such as the Pixel and latest iPhone, have succeeded in greatly reducing low-light sensor noise, but it remains an issue for many.

Existing denoising software, such as Photoshop’s feature or products dedicated to noise removal, do a decent job of removing small amounts of noise but aren’t suitable for very noise images. Artificial intelligence may offer the solution, using deep learning to restore otherwise unusable images to a decent quality. NVIDIA has detailed such a technology.

NVIDIA researchers recently revealed their technology in a new research paper, presenting an AI that can scrub noise from images without having a second noise-free image available. Their system involves the Tensor Flow framework and NVIDIA Tesla P100 GPUs, as well as a dataset containing 50,000 images; it can also remove artifacts like watermarks.

The usefulness extends beyond consumer digital photography and could benefit various industries. For example, the researchers demonstrate their technology being used to clean up a noisy MRI, helping medical professionals discern issues that may otherwise be obscured by the noise. Noisy security camera images could also benefit from the technology.


Must Read Bits & Bytes