T-DDFM: An Advanced Denoising Diffusion Fusion Model for High-Precision Neutron Imaging
T-DDFM: An Advanced Denoising Diffusion Fusion Model for High-Precision Neutron Imaging
Yuan-Yu Liu1
作者信息
- 1. Institute of Energy, Hefei Comprehensive National Science Center (Anhui Energy Laboratory)
- 折叠
摘要
Neutron imaging, a pivotal tool in materials science and nondestructive testing, is frequently hampered bysignificant noise and limited spatial resolution, which obscure critical structural details. This study introducesan advanced denoising diffusion fusion model (DDFM), termed T-DDFM, which incorporates tensor robustprincipal component analysis (TRPCA) for enhanced data preprocessing. This enhancement markedly improvesnoise suppression and image fusion capabilities. When applied to thermal neutron imaging using a 2.5 MeVD-D neutron generator, the T-DDFM algorithm exhibited exceptional performance, detecting structural featuresas small as 0.2 mm in a stainless steel sample 2 cm thick, substantially surpassing the conventional detectionlimit of 0.5 mm. Comprehensive quantitative assessments utilizing key image quality metrics, signal-to-noiseratio (SNR), naturalness image quality evaluator (NIQE), and blind image quality evaluator (BIQI) demonstratethat the T-DDFM consistently outperformed existing methods. It significantly enhances both noise reductionand image fidelity, highlighting its potential for high-precision applications in industrial testing and scientificresearch. This breakthrough positions T-DDFM as a significant advancement in neutron imaging technology,expected to transform practices in fields requiring meticulous material characterization.
Abstract
Neutron imaging, a pivotal tool in materials science and nondestructive testing, is frequently hampered bysignificant noise and limited spatial resolution, which obscure critical structural details. This study introducesan advanced denoising diffusion fusion model (DDFM), termed T-DDFM, which incorporates tensor robustprincipal component analysis (TRPCA) for enhanced data preprocessing. This enhancement markedly improvesnoise suppression and image fusion capabilities. When applied to thermal neutron imaging using a 2.5 MeVD-D neutron generator, the T-DDFM algorithm exhibited exceptional performance, detecting structural featuresas small as 0.2 mm in a stainless steel sample 2 cm thick, substantially surpassing the conventional detectionlimit of 0.5 mm. Comprehensive quantitative assessments utilizing key image quality metrics, signal-to-noiseratio (SNR), naturalness image quality evaluator (NIQE), and blind image quality evaluator (BIQI) demonstratethat the T-DDFM consistently outperformed existing methods. It significantly enhances both noise reductionand image fidelity, highlighting its potential for high-precision applications in industrial testing and scientificresearch. This breakthrough positions T-DDFM as a significant advancement in neutron imaging technology,expected to transform practices in fields requiring meticulous material characterization.关键词
Neutron imaging/ T-DDFM/ noise reduction/ spatial resolution/ nondestructive testing/ image fusionKey words
Neutron imaging/ T-DDFM/ noise reduction/ spatial resolution/ nondestructive testing/ image fusion引用本文复制引用
Yuan-Yu Liu.T-DDFM: An Advanced Denoising Diffusion Fusion Model for High-Precision Neutron Imaging[EB/OL].(2026-07-15)[2026-07-18].https://chinaxiv.org/abs/202607.00116.学科分类
原子能技术应用
