Search arXiv⌕ Search
Home/Archive/arXiv

arXiv · 2004.03145

Plug-and-play ISTA converges with kernel denoisers

Ruturaj G. Gavaskar·Kunal N. Chaudhury
View original source ↗Read source document ↗Export metadata ↓

Abstract

Plug-and-play (PnP) method is a recent paradigm for image regularization, where the proximal operator (associated with some given regularizer) in an iterative algorithm is replaced with a powerful denoiser. Algorithmically, this involves repeated inversion (of the forward model) and denoising until convergence. Remarkably, PnP regularization produces promising results for several restoration applications. However, a fundamental question in this regard is the theoretical convergence of the PnP iterations, since the algorithm is not strictly derived from an optimization framework. This question has been investigated in recent works, but there are still many unresolved problems. For example, it is not known if convergence can be guaranteed if we use generic kernel denoisers (e.g. nonlocal means) within the ISTA framework (PnP-ISTA). We prove that, under reasonable assumptions, fixed-point convergence of PnP-ISTA is indeed guaranteed for linear inverse problems such as deblurring, inpainting and superresolution (the assumptions are verifiable for inpainting). We compare our theoretical findings with existing results, validate them numerically, and explain their practical relevance.

Explore related subjects

eess.IVcs.CV

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Ruturaj G. Gavaskar, Kunal N. Chaudhury. 2020-04-14. Plug-and-play ISTA converges with kernel denoisers. https://doi.org/10.1109/lsp.2020.2986643

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

SONAR: A Structure-Consistent Neural Operator for Null-Space-Aware Sparse View CT Reconstruction

Sparse-view computed tomography (CT) reduces radiation dose and acquisition time but remains severely ill-posed because incomplete projections poorly constrain null-space information. Existing learning-based methods often estimate this information in high-dimensional image space, conflate physical measurement errors with prediction errors, and depend on fixed discretizations. We propose SONAR, a Structure-Consistent Neural Operator for Null-Space-Aware Reconstruction. Instead of recovering the full null-space component, SONAR predicts a low-dimensional null-space-aware representation from the acquired projections as pseudo-measurements. It separates measurement and pseudo-measurement residuals, lifts them into the image domain through physics operators, and applies independent neural operators to constrain their structural effects, thereby accommodating admissible errors while suppressing unsupported structures. To support cross-discretization reconstruction, an anisotropic U-shaped neural operator models the periodic angular and nonperiodic detector dimensions using direction-dependent continuous supports, while image-domain neural operators re-discretize continuous kernels on target grids. These components form an optimization-inspired unrolled network. Experiments on simulated AAPM and clinical MARS photon-counting CT data demonstrate consistent improvements across seen and unseen view settings and unseen image resolutions. On AAPM dataset, SONAR improves PSNR by 1.87~dB at 62 views and by 7.63~dB under zero-shot transfer to a $512\times512$ grid over the strongest competing methods. SONAR also achieves the best overall performance in all clinical settings evaluated, demonstrating accurate, structurally reliable, and discretization-robust sparse-view CT reconstruction.

eess.IV↗
arXiv2026

Physics-informed denoising method for image reconstruction in quantitative low-field MRI

Low-field magnetic resonance imaging (MRI) is becoming increasingly important for medical imaging because it can reduce healthcare costs while ensuring high diagnostic output. Nevertheless, quantitative imaging in low-field MRI faces challenges, such as low signal-to-noise ratio and long scan durations. Deep learning approaches have been proposed for image reconstruction to overcome these challenges. Still, deep learning often requires large high-quality training datasets which are usually not available for low-field applications. Here we propose a modular unrolled end-to-end deep learning method for the denoised reconstruction of quantitative parameter maps directly from k-space data for low-field MRI. It consists of three sub-networks that are iteratively applied. They are used for the regularization of the quantitative parameter estimation, as well as for the signal estimation that is based on simulated signal curves. It generalises well and can be applied to different field strengths and even different quantitative MR sequences without the need for new training data. We applied the presented method to noisy data of knees acquired at 0.55 T for the reconstruction of $T_2$-maps and compared it to other classical and deep learning methods. We also applied the proposed approach to $T_1$-mapping of knees at 72 mT and $T_2$-mapping of brains at 0.6 T. The presented approach outperforms the other reconstruction methods with a median difference below 4 ms to the ground truth $T_2$-map. Even though the network was trained with $T_2$-maps acquired at 0.55 T, it successfully denoised data acquired at different field strengths, sequences, and of different anatomies. As a result, the proposed network and its underlying method offer an efficient and flexible solution to denoise low-field MR data and make quantitative low-field MRI a feasible diagnostic tool for clinical applications.

eess.IV↗
arXiv2026

Recurrent Dynamic Range Extension

We present an approach to progressively extend the highlights of an image. Instead of reconstructing the full dynamic range of a complex scene directly, we learn a simpler task first: We extend the dynamic range of an input image by a single exposure value. Once this is mastered, we retrieve the full HDR image for the scene by executing our network recurrently, progressively increasing the dynamic range of the input. Our formulation is agnostic to the input dynamic range and targets a bounded output domain. This enables us to use widely available RAW images for the reconstruction task and adapt adversarial losses to construct realistic images. By incorporating Memory Replay for backpropagation, we can train our network recurrently over multiple inference stages and reduce reconstruction errors. As a consequence, our system reconstructs challenging long-tailed HDR scenes robustly and shows powerful recovery of bright light sources and highlights.

eess.IV↗