Subscribe to the gold package and get unlimited access to Shamra Academy
Register a new userLearning to Denoise Astronomical Images with U-nets
60
0
0.0
(
0
)
Ask ChatGPT about the research
No Arabic abstract
Astronomical images are essential for exploring and understanding the universe. Optical telescopes capable of deep observations, such as the Hubble Space Telescope, are heavily oversubscribed in the Astronomical Community. Images also often contain additive noise, which makes de-noising a mandatory step in post-processing the data before further data analysis. In order to maximise the efficiency and information gain in the post-processing of astronomical imaging, we turn to machine learning. We propose Astro U-net, a convolutional neural network for image de-noising and enhancement. For a proof-of-concept, we use Hubble space telescope images from WFC3 instrument UVIS with F555W and F606W filters. Our network is able to produce images with noise characteristics as if they are obtained with twice the exposure time, and with minimum bias or information loss. From these images, we are able to recover 95.9% of stars with an average flux error of 2.26%. Furthermore the images have, on average, 1.63 times higher signal-to-noise ratio than the input noisy images, equivalent to the stacking of at least 3 input images, which means a significant reduction in the telescope time needed for future astronomical imaging campaigns.
rate research
Read More
We propose an audio-to-audio neural network model that learns to denoise old music recordings. Our model internally converts its input into a time-frequency representation by means of a short-time Fourier transform (STFT), and processes the resulting complex spectrogram using a convolutional neural network. The network is trained with both reconstruction and adversarial objectives on a synthetic noisy music dataset, which is created by mixing clean music with real noise samples extracted from quiet segments of old recordings. We evaluate our method quantitatively on held-out test examples of the synthetic dataset, and qualitatively by human rating on samples of actual historical recordings. Our results show that the proposed method is effective in removing noise, while preserving the quality and details of the original music.
Generative Adversarial Networks (GANs) are a class of artificial neural network that can produce realistic, but artificial, images that resemble those in a training set. In typical GAN architectures these images are small, but a variant known as Spatial-GANs (SGANs) can generate arbitrarily large images, provided training images exhibit some level of periodicity. Deep extragalactic imaging surveys meet this criteria due to the cosmological tenet of isotropy. Here we train an SGAN to generate images resembling the iconic Hubble Space Telescope eXtreme Deep Field (XDF). We show that the properties of galaxies in generated images have a high level of fidelity with galaxies in the real XDF in terms of abundance, morphology, magnitude distributions and colours. As a demonstration we have generated a 7.6-billion pixel generative deep field spanning 1.45 degrees. The technique can be generalised to any appropriate imaging training set, offering a new purely data-driven approach for producing realistic mock surveys and synthetic data at scale, in astrophysics and beyond.
We present a new method of interpolation for the pixel brightness estimation in astronomical images. Our new method is simple and easily implementable. We show the comparison of this method with the widely used linear interpolation and other interpolation algorithms using one thousand astronomical images obtained from the Sloan Digital Sky Survey. The comparison shows that our method improves bad pixels brightness estimation with four times lower mean error than the presently most popular linear interpolation and has a better performance than any other examined method. The presented idea is flexible and can be also applied to presently used and future interpolation methods. The proposed method is especially useful for large sky surveys image reduction but can be also applied to single image correction.
Astronomical images from optical photometric surveys are typically contaminated with transient artifacts such as cosmic rays, satellite trails and scattered light. We have developed and tested an algorithm that removes these artifacts using a deep, artifact free, static sky coadd image built up through the median combination of point spread function (PSF) homogenized, overlapping single epoch images. Transient artifacts are detected and masked in each single epoch image through comparison with an artifact free, PSF-matched simulated image that is constructed using the PSF-corrected, model fitting catalog from the artifact free coadd image together with the position variable PSF model of the single epoch image. This approach works well not only for cleaning single epoch images with worse seeing than the PSF homogenized coadd, but also the traditionally much more challenging problem of cleaning single epoch images with better seeing. In addition to masking transient artifacts, we have developed an interpolation approach that uses the local PSF and performs well in removing artifacts whose widths are smaller than the PSF full width at half maximum, including cosmic rays, the peaks of saturated stars and bleed trails. We have tested this algorithm on Dark Energy Survey Science Verification data and present performance metrics. More generally, our algorithm can be applied to any survey which images the same part of the sky multiple times.
Purpose: To develop a deep learning approach to de-noise optical coherence tomography (OCT) B-scans of the optic nerve head (ONH). Methods: Volume scans consisting of 97 horizontal B-scans were acquired through the center of the ONH using a commercial OCT device (Spectralis) for both eyes of 20 subjects. For each eye, single-frame (without signal averaging), and multi-frame (75x signal averaging) volume scans were obtained. A custom deep learning network was then designed and trained with 2,328 clean B-scans (multi-frame B-scans), and their corresponding noisy B-scans (clean B-scans + gaussian noise) to de-noise the single-frame B-scans. The performance of the de-noising algorithm was assessed qualitatively, and quantitatively on 1,552 B-scans using the signal to noise ratio (SNR), contrast to noise ratio (CNR), and mean structural similarity index metrics (MSSIM). Results: The proposed algorithm successfully denoised unseen single-frame OCT B-scans. The denoised B-scans were qualitatively similar to their corresponding multi-frame B-scans, with enhanced visibility of the ONH tissues. The mean SNR increased from $4.02 pm 0.68$ dB (single-frame) to $8.14 pm 1.03$ dB (denoised). For all the ONH tissues, the mean CNR increased from $3.50 pm 0.56$ (single-frame) to $7.63 pm 1.81$ (denoised). The MSSIM increased from $0.13 pm 0.02$ (single frame) to $0.65 pm 0.03$ (denoised) when compared with the corresponding multi-frame B-scans. Conclusions: Our deep learning algorithm can denoise a single-frame OCT B-scan of the ONH in under 20 ms, thus offering a framework to obtain superior quality OCT B-scans with reduced scanning times and minimal patient discomfort.
Log in to be able to interact and post comments
comments
Fetching comments
Sign in to be able to follow your search criteria
