Source: University of Illinois Urbana-Champaign
Researchers from the University of Illinois Urbana-Champaign and Intel have developed a post production tool to improve photographs taken in low light scenarios. The researchers used deep-learning techniques to provide a neural network with a data set to enable it to learn the image processing pipeline for images taken in low light situations. The tool uses artificial intelligence to artificially boost the light of photos by an equivalent of up to 300 times the exposure without introducing the levels of discoloration or noise that other tools do. Although the system was trained using photographs taken from high-resolution digital cameras, the development team discovered that the algorithm was also able to improve photos taken from smartphone cameras.
The deep learning system was fed a data set composed of 5,094 short-exposure images that were too dark. The deep learning system was then provided with long-exposure photographs of the same scene, to let the system compare how the scene would look if it had been properly lighted.
Click here to read the full article, and here to view a video demonstration of the tool.
The deep learning system was fed a data set composed of 5,094 short-exposure images that were too dark. The deep learning system was then provided with long-exposure photographs of the same scene, to let the system compare how the scene would look if it had been properly lighted.
Click here to read the full article, and here to view a video demonstration of the tool.
