23 Daily — According to arstechnica.com, in 2026, a research team developed a software package called FLIP (Physics-based Learning for Imaging Planets) that optimizes the physical model of the James Webb Space Telescope’s Near Infrared Camera (NIRCam) to effectively reduce noise and improve the quality of exoplanet imaging.
Software optimizes physical model and reduces imaging noise
When capturing images of exoplanets, the telescope faces challenges because the host stars are much brighter than the planets, and subtle variations in the telescope’s hardware and pointing introduce noise. FLIP models the telescope’s point spread function (PSF), which describes how a telescope blurs a point light source, using a physical hardware model with around one million parameters. The software applies gradient descent optimization—commonly used in training deep neural networks—to fit the model to actual star observation data. Running on modern GPUs, the optimization completes in about half an hour. The model is further fine-tuned on the specific scientific images to account for observation-specific variables, resulting in better noise reduction.
Tests show improved sensitivity and detection of previously unseen exoplanets
The researchers tested FLIP on observations of three stars: two young stars with known exoplanets (HIP 65426 and 51 Eridani) and a third variable star (TYC 5899) with no known planets. The technique increased the signal-to-noise ratio by a factor of 1.5 to 2, depending on observation. Notably, at a wavelength used for detecting carbon dioxide, they were able to identify a planet directly in the Webb data for the first time. For TYC 5899, the sensitivity improvement was up to 15 times. Beyond enhancing images of known planets, FLIP may enable detection of dimmer, older, or lower-mass exoplanets that were previously undetectable.
Editor's take: Software innovation pushes exoplanet imaging frontiers, opening avenues for space technology startups
FLIP demonstrates how advanced physics-based models combined with large-scale parameter optimization can overcome noise limitations in astronomical imaging hardware. For entrepreneurs, this represents a promising approach integrating AI and physical modeling to enhance the performance of scientific instruments. Verification could involve applying FLIP to observations at various wavelengths and different telescopes, assessing image quality improvements and its capability to discover new celestial bodies. Such interdisciplinary software solutions offer innovative entry points into the space technology market.