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

Abstract

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.

A novel imaging method (FIM-ID) reveals that myofibrillogenesis plays a major role in the mechanically induced growth of skeletal muscle

“…Indeed, for first-time users, we suspect that the most complicated part of the workflow will be the implementation of image deconvolution. Fortunately, deconvolution modules have become a commonly available add-on for commercial imaging platforms, and when not available, there are several open-source deconvolution algorithms that users could deploy ( Lam et al, 2017 ; Makarkin and Bratashov, 2021 ; Katoh, 2019 ; Su et al, 2023 ). As such, we are confident that a broad range of investigators will be able to take advantage of our method.…”

Fine‐Grained Scene Graph Generation with Overlap Region and Geometrical Center

“…Deconvolution : Deconvolution is known as transposed convolution, but not the reverse operation of convolution [YYT*21]. Deconvolution can be used to solve the problem that the feature map becomes smaller after a series of previous convolution operations [MB21,APS19]. It can be used to convert a small feature map into a large feature map and obtain the relative position relationship information in the feature map.…”

MR‐zero meets RARE MRI: Joint optimization of refocusing flip angles and neural networks to minimize T2‐induced blurring in spin echo sequences

“…Blurring is a major cause of image degradation and corrupts valuable image information 4 . In the image processing and computer vision community, deblurring is a traditional inverse problem with the aim to recover a sharp image from the corresponding degraded image 5 . The blurring problem can be classified in two types, non‐blind deblurring problems and blind deblurring problems 6 .…”

“…4 In the image processing and computer vision community, deblurring is a traditional inverse problem with the aim to recover a sharp image from the corresponding degraded image. 5 The blurring problem can be classified in two types, non-blind deblurring problems and blind deblurring problems. 6 The former case, non-blind deblurring, indicates that the blur kernel is assumed to be known and a sharp image can be induced from both the blurry image and the kernel.…”