Boffins use machine-learning tricks to create a starlight dimmer switch

Scientists have developed a method that removes a star’s glare from telescope images, making faint exoplanets easier to detect. Tested on James Webb Space Telescope data, the approach revealed a known planet in an image where conventional processing had failed, and could help improve the search for planets around other stars.

The framework, called Physics-based Learning for Imaging Planets (FLIP), combines a model of how light travels through a telescope with machine-learning optimisation to account for small optical imperfections. In tests, it improved sensitivity by between two and 15 times; it also produced the first image of 51 Eridani b at an infrared wavelength of 4.3 micrometres. The researchers say further work must reduce computing demands and handle larger ground-based datasets.