Finite-resolution Deconvolution of Multiwavelength Imaging of 20,000 Galaxies in the COSMOS Field: The Evolution of Clumpy Galaxies over Cosmic Time

Compact star-forming clumps observed in distant galaxies are often suggested to play a crucial role in galaxy assembly. In this paper, we use a novel approach of applying finite-resolution deconvolution on ground-based images of the COSMOS field to resolve 20,185 star-forming galaxies (SFGs) at 0.5 < z < 2 to an angular resolution of 0.″3 and study their clump fractions. A comparison between the deconvolved images and HST images across four different filters shows good agreement and validates image deconvolution. We model spectral energy distributions using the deconvolved 14-band images to provide resolved surface brightness and stellar-mass density maps for these galaxies. We find that the fraction of clumpy galaxies decreases with increasing stellar masses and with increasing redshift: from ∼30% at z ∼ 0.7 to ∼50% at z ∼ 1.7. Using abundance matching, we also trace the progenitors for galaxies at z ∼ 0.7 and measure the fractional mass contribution of clumps toward their total mass budget. Clumps are observed to have a higher fractional mass contribution toward galaxies at higher redshift: increasing from ∼1% at z ∼ 0.7 to ∼5% at z ∼ 1.7. Finally, the majority of clumpy SFGs have higher specific star formation rates (sSFR) compared to the average SFGs at fixed stellar mass. We discuss the implication of this result for in situ clump formation due to disk instability.

State-of-the-Art Approaches for Image Deconvolution Problems, including Modern Deep Learning Architectures

In modern digital microscopy, deconvolution methods are widely used to eliminate a number of image defects and increase resolution. In this review, we have divided these methods into classical, deep learning-based, and optimization-based methods. The review describes the major architectures of neural networks, such as convolutional and generative adversarial networks, autoencoders, various forms of recurrent networks, and the attention mechanism used for the deconvolution problem. Special attention is paid to deep learning as the most powerful and flexible modern approach. The review describes the major architectures of neural networks used for the deconvolution problem. We describe the difficulties in their application, such as the discrepancy between the standard loss functions and the visual content and the heterogeneity of the images. Next, we examine how to deal with this by introducing new loss functions, multiscale learning, and prior knowledge of visual content. In conclusion, a review of promising directions and further development of deconvolution methods in microscopy is given.