An artificial intelligence algorithm “blurs” the universe

Images taken by ground-based telescopes appear hazy due to Earth’s atmosphere. So a team of scientists used an artificial intelligence algorithm that removes blur and provides more accurate images. The research results have been published in the Royal Astronomical Society’s journal Monthly Notices.

Even images taken by the best ground-based telescopes are blurred due to shifting pockets of air in the atmosphere, and this noise can lead to incorrect measurements.

Scientists from Northwestern University in the US and Tsinghua University in China have presented a new strategy to solve the problem.

They have adapted a machine vision algorithm commonly used for image smoothing to improve noise in images captured by ground-based telescopes.

They also trained the algorithm on simulated data to match the imaging parameters of the Vera C. Rubin Observatory in Chile so that when it opens next year, the instrument will be in direct compliance.

A simulated image with atmospheric blurring, in which the researchers used images from the Hubble Space Telescope and added atmospheric effects (Credit: Emma Alexander/Northwestern University)

The researchers then used an artificial intelligence algorithm to de-blur the atmosphere, revealing the real image (Credit: Emma Alexander/Northwestern University, APE – MPE).

The tool runs faster and produces more realistic images than current technologies. With it, the images had 38.6% fewer errors than traditional deinterlacing methods and 7.4% fewer errors than modern methods.

“By cleaning the images the right way, we can get more accurate data. The algorithm computationally removes the atmosphere allowing physicists to get better scientific measurements. At the end of the day, Emma Alexander, assistant professor of computer science at Northwestern University and one of the study’s authors, says: The pictures look better too.”

For astronomers interested in using this instrument, open source software and instructions are available at www https://github.com/Lukeli0425/Galaxy-Deconv.