Imaging Signal Recovery Using Neural Network Priors Under Uncertain Forward Model Parameters

Xiwen Chen,Wenhui Zhu,Peijie Qiu,Abolfazl Razi

2024-05-05

Abstract:Inverse imaging problems (IIPs) arise in various applications, with the main objective of reconstructing an image from its compressed measurements. This problem is often ill-posed for being under-determined with multiple interchangeably consistent solutions. The best solution inherently depends on prior knowledge or assumptions, such as the sparsity of the image. Furthermore, the reconstruction process for most IIPs relies significantly on the imaging (i.e. forward model) parameters, which might not be fully known, or the measurement device may undergo calibration drifts. These uncertainties in the forward model create substantial challenges, where inaccurate reconstructions usually happen when the postulated parameters of the forward model do not fully match the actual ones. In this work, we devoted to tackling accurate reconstruction under the context of a set of possible forward model parameters that exist. Here, we propose a novel Moment-Aggregation (MA) framework that is compatible with the popular IIP solution by using a neural network prior. Specifically, our method can reconstruct the signal by considering all candidate parameters of the forward model simultaneously during the update of the neural network. We theoretically demonstrate the convergence of the MA framework, which has a similar complexity with reconstruction under the known forward model parameters. Proof-of-concept experiments demonstrate that the proposed MA achieves performance comparable to the forward model with the known precise parameter in reconstruction across both compressive sensing and phase retrieval applications, with a PSNR gap of 0.17 to 1.94 over various datasets, including MNIST, X-ray, Glas, and MoNuseg. This highlights our method's significant potential in reconstruction under an uncertain forward model.

Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

### Problems Addressed by the Paper The paper primarily addresses the challenge of image reconstruction in Inverse Imaging Problems (IIPs) caused by the uncertainty of forward model parameters. Specifically: 1. **Uncertainty of Forward Model Parameters**: - In many practical applications, such as Magnetic Resonance Imaging (MRI) and tomography, the forward model parameters may not be fully known, or the equipment may experience calibration drift, leading to increased uncertainty in the reconstruction process. - These uncertainties make it very difficult to reconstruct the original image from compressed measurements. 2. **Proposed New Method**: - The paper proposes a new framework—**Moment-Aggregation (MA)**, which can perform signal reconstruction in the presence of a set of candidate forward model parameters. - The MA framework constructs a loss function by simultaneously considering the impact of all candidate parameters and utilizes the prior knowledge of neural networks for optimization. - Experiments demonstrate that this method performs comparably to when the exact parameters are known, with PSNR differences ranging from 0.17 to 1.94. ### Main Contributions 1. Proposed the Moment-Aggregation framework to address inverse imaging problems under uncertain forward model parameters. 2. Theoretically analyzed the convergence of the MA framework and experimentally validated its effectiveness. This research has significant application potential in fields such as medical imaging, particularly in achieving high-quality image reconstruction under inaccurate parameter conditions.

Imaging Signal Recovery Using Neural Network Priors Under Uncertain Forward Model Parameters

A Set-Theoretic Study of the Relationships of Image Models and Priors for Restoration Problems

Sparse-view Signal-domain Photoacoustic Tomography Reconstruction Method Based on Neural Representation

Implicit Neural Representation for Sparse-view Photoacoustic Computed Tomography

Deep plug-and-play MRI reconstruction based on multiple complementary priors

Implicit Neural Representation for MRI Parallel Imaging Reconstruction

High-dimensional embedding network derived prior for compressive sensing MRI reconstruction

A Symmetric Forward-Inverse Reinforcement Framework for Image Reconstruction Through Scattering Media

Multiply Complementary Priors for Image Compressive Sensing Reconstruction in Impulsive Noise

NPB-REC: A Non-parametric Bayesian Deep-learning Approach for Undersampled MRI Reconstruction with Uncertainty Estimation

Image Reconstruction without Explicit Priors

Ghost Imaging by Single-Layer Neural Network Based on Forward Physical Model

Untrained Neural Network Priors for Inverse Imaging Problems: A Survey

Subspace Model-Assisted Deep Learning for Improved Image Reconstruction.

Uncertainty-Aware Null Space Networks for Data-Consistent Image Reconstruction

Joint learning of nonlinear representation and projection for fast constrained MRSI reconstruction

Highly Accelerated MRI via Implicit Neural Representation Guided Posterior Sampling of Diffusion Models

BlindNet: an untrained learning approach toward computational imaging with model uncertainty

A Unified Model for Compressed Sensing MRI Across Undersampling Patterns

Analysis of Deep Image Prior and Exploiting Self-Guidance for Image Reconstruction

Accelerated model‐based iterative reconstruction strategy for sparse‐view photoacoustic tomography aided by multi‐channel autoencoder priors