An interdisciplinary team from Tsinghua University in China has developed an artificial intelligence model aimed at enhancing astronomical imaging and pushing the boundaries of deep space exploration. Named ASTERIS (Astronomical Spatiotemporal Enhancement and Reconstruction for Image Synthesis), this system combines computational optics and AI algorithms.
The results, published in the journal Science, show that the model can extract extremely faint astronomical signals, identify galaxies located more than 13 billion light-years away, and produce the deepest images of the cosmos ever obtained.
Observing very distant celestial objects is essential to understanding the origin and evolution of the universe. However, astronomers face a major obstacle: the weak signals from these objects are often masked by sky background noise and thermal radiation from instruments.
ASTERIS introduces a technique called « spatiotemporal self-supervised denoising. » Applied to data from the James Webb Space Telescope (JWST), it extends the observation coverage from the visible spectrum (around 500 nanometers) to the mid-infrared (5 micrometers). The model also increases detection depth by one magnitude, allowing identification of objects 2.5 times fainter than before.
Thanks to this technology, the team identified more than 160 candidate high-redshift galaxies dating from the so-called « cosmic dawn » period, between 200 and 500 million years after the Big Bang. This number represents a threefold increase compared to previous methods, according to Cai Zheng, associate professor in the Department of Astronomy at Tsinghua.
Unlike traditional noise reduction techniques, which assume uniform interference, ASTERIS reconstructs images as a three-dimensional spatiotemporal volume. Through an adaptive photometric analysis mechanism, it distinguishes noise fluctuations from the extremely faint signals emitted by distant stars and galaxies.
Researchers estimate that the model can process enormous volumes of data from different telescopes, potentially making it a universal platform for enhancing space data. The goal is to integrate it into future generations of telescopes to deepen the study of dark energy, dark matter, cosmic origins, and exoplanets.
Experts believe this advancement could have a major impact on modern astronomy by improving the ability to detect and analyze the oldest and most distant objects in the universe.
