[ad_1]
NVIDIA researchers are back with another digital imaging technology that pushes the boundaries of traditional image manipulation. Unlike the project recently unveiled by Adobe, which involved a neural network trained to spot digitally altered images, NVIDIA's latest creation can clean digital noise from sensors and watermarks from digital images. Thanks to the artificial intelligence that powers it, the functionality is much more efficient than existing denoising tools.
The sound of the digital camera sensor, although less severe, is still prevalent in mainstream cameras, especially light conditions This is due to the small size of sensor used in these cameras, making the post-processing necessary to improve the quality of the image. Some products, such as the Pixel and the latest iPhone, have been able to significantly reduce the noise of low-light sensors, but this remains a problem for many.
Existing denoising software, such as the function of Photoshop or products dedicated to noise, decent work to remove small amounts of noise, but are not suitable for very noise images. Artificial intelligence can offer the solution, using in-depth learning to restore otherwise unusable images to decent quality. NVIDIA has detailed such technology.
NVIDIA researchers recently unveiled their technology in a new research paper, presenting an AI capable of cleaning the sound of images without having a second image without noise. Their system includes the Tensor Flow framework and the NVIDIA Tesla P100 GPUs, as well as a dataset containing 50,000 images; it can also remove artifacts such as watermarks.
The utility extends beyond mainstream digital photography and could benefit various industries. For example, researchers show that their technology is used to clean a noisy MRI, helping health professionals to discern problems that might otherwise be obscured by noise. Images of noisy security cameras could also benefit from this technology.
SOURCE: NVIDIA
Source link
