Researchers from Intel and the University of Illinois Urbana-Champlain (UIUC) have developed a new procedure that is intended to solve a common problem with many digital cameras, including the latest models: As the tiny sensors can capture only a limited amount of light, the images are often too dark and can also be blurred. In their research report , the scientists describe how they fed an AI algorithm with data from photos with insufficient light, then recorded the images again with longer exposure times and displayed them on the computer. They processed a total of 5,000 photos in this way for training purposes.
The algorithm is supposed to be able to brighten dark images by a factor of 300 without causing the dreaded image noise and color changes that come about when processing with programs such as Photoshop. And even though the training images were taken with a high-resolution digital camera, the procedure also worked in a test with an iPhone 6. The code and database for the project have now been published on GitHub .