A U-net trained on noisy image pairs alone, without clean targets, denoises solar Stokes images to about 6e-4 continuum residual, matching clean-target training on synthetic data.
Removal of Spectro-Polarimetric Fringes by 2D PCA
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
abstract
We investigate the application of 2-dimensional Principal Component Analysis (2D PCA) to the problem of removal of polarization fringes from spectro-polarimetric data sets. We show how the transformation of the PCA basis through a series of carefully chosen rotations allows to confine polarization fringes (and other stationary instrumental effects) to a reduced set of basis "vectors", which at the same time are largely devoid of the spectral signal from the observed target. It is possible to devise algorithms for the determination of the optimal series of rotations of the PCA basis, thus opening the possibility of automating the procedure of de-fringing of spectro-polarimetric data sets. We compare the performance of the proposed method with the more traditional Fourier filtering of Stokes spectra.
fields
astro-ph.SR 1years
2019 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Solar image denoising with convolutional neural networks
A U-net trained on noisy image pairs alone, without clean targets, denoises solar Stokes images to about 6e-4 continuum residual, matching clean-target training on synthetic data.