Skip to content
Preprint

Marginalization-Based Regularization for Separable Nonlinear Least Squares Problems with Applications to Semi-Blind Deblurring

Nov 2026 · 0 citations · 28 references
Mathematics Computer Science

TL;DR

A family of blur-parameter priors whose weight and auxiliary operators can be chosen independently of the image prior are constructed, motivated by the determinant term arising from Gaussian marginalization.

Abstract

Separable nonlinear least squares problems arise in inverse problems such as semi-blind image deblurring, where the image enters the forward model linearly, and the blur parameters enter nonlinearly. Even when the image is regularized, joint estimation of the image and blur parameters can favor a degenerate no-blur solution, in which the blur operator tends to the identity and the reconstruction merely denoises the data without removing the blur. We address this by regularizing the blur parameters. Motivated by the determinant term arising from Gaussian marginalization, we construct a family of blur-parameter priors whose weight and auxiliary operators can be chosen independently of the image prior. The resulting log-determinant penalty is incorporated into regularized variable projection, which eliminates the image variables by minimization. Under periodic boundary conditions and assuming the matrices involved admit a joint spectral decomposition, we show that a sufficiently large determinant weight rules out the no-blur solution as a local minimizer. We also characterize the regularizer's behavior as the auxiliary regularization scale tends to zero or infinity. We solve the reduced problem with a quasi-Newton method that evaluates the objective and its derivatives in the Fourier domain. In experiments with Gaussian, defocus, and motion blur, the proposed prior improves image and blur recovery over Gaussian marginalization across several settings, with determinant weights computed from the true image. A single-sample experiment illustrates how to automatically choose parameters from the data using a conditional generalized cross-validation criterion.

View source

Similar papers

Oct 2026

Accelerated stochastic iterative algorithms for image reconstruction via enhanced Regularization-by-Denoising

Inverse imaging problems such as denoising and deblurring aim to recover a clean image from corrupted measurements by incorporating prior knowledge. Classical regularizers like Tikhonov and total variation are mathematically convenient, yet they often oversimplify the complex structure of natural images. Regularization...

Yasmine El Mobariki, Brahim Khaider, Amine Laghrib et al. · 0 citations
Oct 2026

Image Restoration with Multiplicative Noise: A Hybrid Geometric Model and Its Alternating Minimization Algorithm

A hybrid regularization is introduced that combines these two complementary quantities to mitigate their respective drawbacks and provides a balanced control over both area and curvature, thereby preserving the advantages of each component.

Sheng-Kun Yang, Zhi-Chang Guo, Yao Li et al. · 1 citation
Conference Sep 2026

Blind Single-Image Restoration and Super-Resolution with Inverse PINNs and Trainable PDE Priors

Single-image restoration and super-resolution are inherently ill-posed inverse problems in which the degradation operator is rarely known at inference time. Supervised feedforward networks address this ambiguity by learning a prior from large paired datasets, but their performance degrades when the actual degradation d...

Alessandro Bottino, M. Lapegna, Andrea Salvatore et al. · 0 citations
Preprint Nov 2026

A Weighted Second-Order Nonlocal Total Variation Regularizer for Image Restoration

Despite the remarkable progress of deep learning in image processing, variational methods remain a fundamental and actively developing framework, offering interpretability, mathematical rigor, and well-founded guarantees. In this work, we introduce a weighted second-order nonlocal total variation regularizer for image...

Francesc Alcover, Joan Duran, Catalina Sbert · 0 citations
Sep 2026

Three-dimensional blind deconvolution by CP-parameterized kernels

We consider the problem of reconstructing a kernel $k$ and image $u$ in a blind deconvolution problem from convolutional data $g = k*u$. We particularly focus on the 3D case, i.e., on volumetric images. The problem is approached by imposing a semiparametric CP decomposition for the kernel and a variational TV penalty f...

Fatoumata Sanogo, S. Kindermann · 0 citations
Preprint Sep 2026

Orlicz space relaxation of total variation denoising

We consider variational image denoising with spatially dependent Orlicz regularization and introduce an Orlicz--Sobolev approximation of the ROF model based on a scaled $L\log L$-type density. For a general class of uniformly superlinear integrands satisfying a $\Delta_2$-condition, we establish existence and uniquenes...

Christian Clason, Tobias Unterberger · 0 citations

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.