DiffNR: Diffusion-Enhanced Neural Representation Optimization for Sparse-View 3D Tomographic Reconstruction

TL;DR

DiffNR introduces a diffusion-based framework with SliceFixer to improve sparse-view 3D CT reconstruction, reducing artifacts and enhancing image quality efficiently.

Contribution

The paper presents DiffNR, a novel diffusion-enhanced neural representation method with SliceFixer for artifact correction in sparse-view CT reconstruction.

Findings

DiffNR improves PSNR by 3.99 dB on average.

The method generalizes well across different domains.

It achieves better runtime performance than iterative denoising methods.

Abstract

Neural representations (NRs), such as neural fields and 3D Gaussians, effectively model volumetric data in computed tomography (CT) but suffer from severe artifacts under sparse-view settings. To address this, we propose DiffNR, a novel framework that enhances NR optimization with diffusion priors. At its core is SliceFixer, a single-step diffusion model designed to correct artifacts in degraded slices. We integrate specialized conditioning layers into the network and develop tailored data curation strategies to support model finetuning. During reconstruction, SliceFixer periodically generates pseudo-reference volumes, providing auxiliary 3D perceptual supervision to fix underconstrained regions. Compared to prior methods that embed CT solvers into time-consuming iterative denoising, our repair-and-augment strategy avoids frequent diffusion model queries, leading to better runtime performance. Extensive experiments show that DiffNR improves PSNR by 3.99 dB on average, generalizes well across domains, and maintains efficient optimization.