New software sharpens Webb Telescope's exoplanet images

Researchers have developed new software called FLIP that significantly reduces noise in exoplanet imaging from the James Webb Space Telescope, boosting signal-to-noise ratios by 1.5 to 2 times and enabling first-time detection of a planet at a carbon dioxide-sensitive wavelength. The software optimizes a physical model of the telescope's NIRCam instrument with 106 parameters using gradient descent, a technique common in AI training. It starts with a simulated point spread function, then fine-tunes it against real star data and the specific scientific image being analyzed, running in about half an hour on a GPU. This approach improves on traditional methods by accounting for subtle telescope variations between observations. Beyond sharper views of known planets, the reduced noise could reveal dimmer, older, or lower-mass exoplanets previously undetected, and the team suggests FLIP could be adapted to identify hardware changes.