How AI can make galactic telescope images ‘sharper’

Accuracy is everything when studying deep space, and this open-source AI is here to help.
Comparison images of galaxy gaining better resolution via AI program
Before and after, all thanks to AI clarification. Emma Alexander/Northwestern University>

Even the most advanced ground-based telescopes struggle with nearsighted vision issues. Often this isn’t through any fault of their own, but a dilemma of having to see through the Earth’s constantly varying atmospheric interferences. As undesirable as that is to the casual viewer, it can dramatically frustrate researchers’ abilities to construct accurate images of the universe—both literally and figuratively. By applying an existing, open-source computer vision AI algorithm to telescope tech, however, researchers have found they are able to hone our cosmic observations.

As detailed in a paper published this month with the Monthly Notices of the Royal Astronomical Society, a team of scientists from Northwestern University and Beijing’s Tsinghua University recently trained an AI on data simulated to match imaging parameters for the soon-to-be opened Vera C. Rubin Observatory in north-central Chile. As Northwestern’s announcement explains, while similar technology already exists, the new algorithm produces blur-free, high resolution glimpses of the universe both faster and more realistically.

“Photography’s goal is often to get a pretty, nice-looking image. But astronomical images are used for science,” said Emma Alexander, an assistant professor of computer science at Northwestern and the study’s senior author. Alexander explained that cleaning up image data correctly helps astronomers obtain far more accurate data. Because the AI algorithm does so computationally, physicists can glean better measurements.

[Related: The most awesome aerospace innovations of 2022.]

The results aren’t just prettier galactic portraits, but more reliable sources of study. For example, analyzing galaxies’ shapes can help determine gravitational effects on some of the universe’s largest bodies. Blurring that image—be it through low-resolution tech or atmospheric interference—makes scientists’ less reliable and accurate. According to the team’s work, the optimized tool generated images with roughly 38 percent less error than compared to classic blur-removal methods, and around 7 percent less error compared to existing modern methods.

What’s more, the team’s AI tool, coding, and tutorial guidelines are already available online for free. Going forward, any interested astronomers can download and utilize the algorithm to improve their own observatories’ telescopes, and thus obtain better and more accurate data.

“Now we pass off this tool, putting it into the hands of astronomy experts,” continued Alexander. “We think this could be a valuable resource for sky surveys to obtain the most realistic data possible.” Until then, astronomy fans can expect far more detailed results from the Rubin Observatory when it officially opens in 2024 to begin its deep survey of the stars.