Select Language:
Researchers in China have created an innovative artificial intelligence model aimed at advancing astronomical imaging, allowing scientists to see deeper into the universe than ever before.
A multidisciplinary team from Tsinghua University developed the model, called ASTERIS (Astronomical Spatiotemporal Enhancement and Reconstruction for Image Synthesis), by leveraging cutting-edge computational optics and AI algorithms.
As reported in the journal Science on February 20, the model enables the extraction of extremely faint astronomical signals, facilitating the identification of galaxies over 13 billion light-years away and producing the most detailed images of deep space to date.
Studying distant and dim celestial bodies is vital for understanding the universe’s origin and evolution. However, astronomers often struggle with weak signals from remote objects, which are frequently obscured by background sky noise and thermal emissions from telescopes.
The research illustrates that applying ASTERIS’s “self-supervised spatiotemporal denoising” technique to data from the James Webb Space Telescope (JWST) extends the observable range from visible light at roughly 500 nanometers to mid-infrared wavelengths at 5 micrometers. This approach also improves detection capabilities by 1.0 magnitude, meaning the telescope can now detect objects 2.5 times fainter than before.
Using this model, the team discovered over 160 potential high-redshift galaxies from the “Cosmic Dawn” era, about 200 to 500 million years after the Big Bang—tripling the number of discoveries made with previous methods, noted Cai Zheng, an associate professor at Tsinghua’s Department of Astronomy.
The researchers believe that this AI model can process huge datasets from space telescopes and is compatible across various observational platforms, positioning it as a potential universal tool for enhancing deep-space data.
Traditional noise reduction methods often involve stacking multiple images and assume noise is uniform or correlated, but in reality, the noise in deep space varies quite a bit over time and location. ASTERIS tackles this challenge by reconstructing images as a three-dimensional spatiotemporal volume.
The model’s “photometric adaptive screening mechanism” can identify subtle fluctuations in noise, effectively differentiating between background interference and the faint signals from distant galaxies and stars.
One expert involved in the research remarked that this work has significant implications for astronomy as a whole.
Dai Qionghai, a professor at Tsinghua University’s Department of Automation, explained that faint celestial objects can now be reconstructed with high accuracy, even when obscured by light noise.
Looking forward, scientists anticipate deploying this groundbreaking technology on future telescopes to explore fundamental questions about dark energy, dark matter, the origins of the universe, and exoplanets.
ChatGPT
Perplexity
Gemini AI
Grok AI
