Study on the Influence of Image Denoising on the Quantitative Characterization of Pore Structure Parameters of Digital Core

Accurate characterization of pore‐structure parameters in digital rock images is critical for the reliable prediction of subsurface transport processes, including fluid flow in hydrocarbon reservoirs and geological CO2 storage formations. X‐ray computed tomography (XCT) enables nondestructive 3D core imaging, but imaging noise degrades reconstruction accuracy and pore network authenticity. Existing research lacks in‐depth analysis of denoising methods’ correlation with pore parameters. This study evaluates XCT datasets of carbonate cores, using entropy, structural similarity index (SSIM), porosity, and fractal dimension to compare traditional denoising methods via 3D pore network modeling. Findings show denoising methods significantly affect porosity calculations: total and connected porosity peak relative differences reach 30.5% and 40.95%, respectively. Tortuosity and pore coordination number are notably impacted, whereas fractal dimension and equivalent pore radius are less affected. Dual‐filter combinations outperform single filters, whereas combinations of three or more filters are not recommended due to excessive smoothing. Although denoising alters parameter values, it preserves the spatial variation trend of core properties, guiding researchers to balance accuracy and efficiency when selecting denoising strategies. This study provides theoretical guidance for scientific denoising algorithm selection in digital core construction, enhancing microscopic reservoir characterization credibility and promoting precise digital core applications in petroleum engineering.