Instrument-To-Instrument translation: Instrumental advances drive restoration of solar observation series via deep learning

Abstract

The constant improvement of astronomical instrumentation provides the foundation for scientific discoveries. In general, these improvements have only implications forward in time, while previous observations do not benefit from this trend. Here we provide a general deep learning method that translates between image domains of different instruments (Instrument-To-Instrument translation; ITI). We demonstrate that the available data sets can directly profit from the most recent instrumental improvements, by applying our method to five different applications of ground- and space-based solar observations. We obtain 1) solar full-disk observations with unprecedented spatial resolution, 2) a homogeneous data series of 24 years of space-based observations of the solar EUV corona and magnetic field, 3) real-time mitigation of atmospheric degradations in ground-based observations, 4) a uniform series of ground-based H$\alpha$ observations starting from 1973, 5) magnetic field estimates from the solar far-side based on EUV imagery. The direct comparison to simultaneous high-quality observations shows that our method produces images that are perceptually similar and match the reference image distribution.

References in corpus (29)

• Instance Normalization: The Missing Ingredient for Fast Stylization
• The Solar Optical Telescope for the Hinode Mission: An Overview
• Deep Learning for Single Image Super-Resolution: A Brief Review
• The Perception-Distortion Tradeoff
• Generative Adversarial Networks recover features in astrophysical images of galaxies beyond the deconvolution limit
• Diversity-Sensitive Conditional Generative Adversarial Networks
• Photometric and Thermal Cross-calibration of Solar EUV Instruments
• A Machine Learning Dataset Prepared From the NASA Solar Dynamics Observatory Mission
• Enhancing SDO/HMI images using deep learning
• Analysis of full disc Ca II K spectroheliograms. II. Towards an accurate assessment of long-term variations in plage areas
• Real-time multiframe blind deconvolution of solar images
• Fast Point Spread Function Modeling with Deep Learning
• Point spread function of SDO/HMI and the effects of stray light correction on the apparent properties of solar surface phenomena
• Real-time flare detection in ground-based H$\alpha$ imaging at Kanzelhöhe Observatory
• Fast Solar Image Classification Using Deep Learning and its Importance for Automation in Solar Physics
• Solar Coronal Magnetic Field Extrapolation from Synchronic Data with AI-generated Farside
• Solar image denoising with convolutional neural networks
• Multi-channel coronal hole detection with convolutional neural networks
• Counter-streaming flows in a giant quiet-Sun filament observed in the extreme ultraviolet
• Kanzelhöhe Observatory: instruments, data processing and data products
• Solar Image Restoration with the Cycle-GAN Based on Multi-Fractal Properties of Texture Features
• New Homogeneous Dataset of Solar EUV synoptic maps from SOHO/EIT and SDO/AIA
• Learning to do multiframe wavefront sensing unsupervisedly: applications to blind deconvolution
• Multi-Channel Auto-Calibration for the Atmospheric Imaging Assembly using Machine Learning
• Exploring the Limits of Synthetic Creation of Solar EUV Images via Image-to-Image Translation
• Image Quality Assessment for Full-Disk Solar Observations with Generative Adversarial Networks
• SuNeRF: 3D reconstruction of the solar EUV corona using Neural Radiance Fields
• Accelerating Multiframe Blind Deconvolution via Deep Learning
• Single-Frame Super-Resolution of Solar Magnetograms: Investigating Physics-Based Metrics & Losses