A Dual CT Image Denoising Approach Using Guided Filter and Method-Based Noise in the NSST Domain

Background This study focuses on Gaussian noise in CT images, which obscures details and hinders interpretation. Electronic interference and environmental factors often generate this type of noise during image acquisition. Therefore, effective denoising is crucial for improved imaging and diagnostic precision. Aim The main aim of this research is to improve the quality of Low-Dose Computed Tomography (LDCT) imaging by suppressing Gaussian noise, preserving edges and sharp features, and improving visual quality and clarity. This, in turn, will enhance the diagnostic precision. Objective The primary goal of this method is to effectively denoise CT images affected by Gaussian noise. It outperforms existing denoising techniques, improving Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Iidex Measure (SSIM), Entropy Difference (ED), Feature Similarity Index measure (FSIM), and Root mean squared error (RMSE), thereby improving overall visual quality. Methods This study evaluates a denoising method combining Nonsubsampled Shearlet Transform (NSST), guided filtering, and BayesShrink thresholding for CT images, achieving superior noise reduction and edge preservation. It was compared with Discrete Wavelet Transform (DWT), NSST with bilateral filtering, Method noise-based Convolutional Neural Network (CNN), NSST with Bayes shrinkage, NSST with Wiener filtering, and Stein’s Unbiased Risk Estimate Linear Expansion of Thresholds (SURELET), and Tetrolet transform. Quantitative evaluation (PSNR, SSIM, ED, FSIM, and RMSE) across noise levels (σ = 5,10,15,20) confirms its consistent superiority in noise minimization and edge preservation. Results The results confirm the proposed approach is superior to other methods in terms of PSNR, SSIM, ED, FSIM and RMSE to improve overall visual quality. Conclusion This proposed hybrid approach, which combines NSST, Bayesian thresholding, guided filtering, and method noise-based approach, effectively reduces Gaussian noise while preserving edges and structural details.