ASTRA-SR: Atmospheric Seeing and Turbulence Restoration for Astronomical Image Super-Resolution
In the authors' words
Ground-based planetary imaging suffers from atmospheric turbulence, sensor noise, and limited sampling, making restoration a joint denoising, deblurring, and super-resolution problem. We present ASTRA-SR, a blind single-frame restoration framework trained on a physics-grounded synthetic dataset. High-dynamic-range spacecraft RAW observations serve as clean sources, and paired LR inputs are synthesized using measured layer-integrated turbulence strengths, propagated moving phase screens, exposure-averaged spatially varying PSFs, and sensor noise.ASTRA-SR first estimates a noise-suppressed but blur-retaining LR image, then restores spatial structure through multiscale processing and reconstructs HR detail with serial spatial-amplitude refinement. It yields a 0.49 dB foreground PSNR gain over the strongest baseline approaches.
Appeared: Wednesday, September 23. arXiv. Preprint, not yet peer-reviewed.
Authors' comment: 4 pages of main text plus references, 4 figures. Submitted to ICASSP 2027