The neural network was used to clean the space images from interference.

Scientists from Northwestern University in the US and Tsinghua University in Beijing created a neural network that was used to clean images of space from interference. The conclusions of the work are published in the arXiv preprints repository. Pictures taken by the world’s best ground-based telescopes usually appear blurry due to pockets of moving air in the atmosphere. This is because light from distant stars, planets, and galaxies passes through Earth’s atmosphere. The atmosphere not only blocks certain light waves, but also distorts them. Even in a clear night sky, the stars twinkle due to vibrations in the air.

This interference obscures the shapes of objects in astronomical images, leading to erroneous physical measurements to study the universe. Therefore, an elliptical galaxy may appear more round or stretched out than it actually is. To solve the problem, the experts modified the artificial intelligence algorithm used to sharpen photos and combined it with a deep learning program. Astrophysicists use similar tools to remove blur, but the new program is faster and produces more realistic and beautiful images. As a result, the neural network cleaned up images that had 38.6% less distortion than traditional blurring methods. In addition, the program has an open source code and an online guide accessible to all astronomers who want to use the development.