A Soft-Threshold Filtering Approach for Tomography Reconstruction from a Limited Number of Projections with Bilateral...

Area of Science:

• Medical Imaging
• Computational Imaging
• Image Reconstruction

Background:

• Statistical methods with L2 norm fidelity and Lp norm regularization (1 < p < 2) are successful in X-ray tomography.
• Discrete Gradient Transform (DGT) sparse transformation with Total Variation (TV) minimization offers good results and computational performance.

Purpose of the Study:

• To improve X-ray tomography image reconstruction by introducing a novel regularization term.
• To enhance noise reduction and image quality, particularly in low-dose reconstruction scenarios.

Main Methods:

• Incorporation of a bilateral edge-preserving (BEP) regularization term into the objective function.
• Alternating minimization of an objective function combining L2 fidelity and the proposed BEP regularization with DGT-TV.

Main Results:

• The addition of BEP regularization adaptively reduces noise in the initial reconstruction phase.
• Improved optimization of the fidelity term and enhanced results for DGT-TV minimization.
• Higher Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index Measurement (SSIM) in low-dose reconstructions.

Conclusions:

• The proposed BEP regularization method effectively improves X-ray tomography image reconstruction.
• BEP enhances noise control, leading to superior image quality and quantitative metrics in challenging low-dose conditions.