A novel grid regression demodulation method for radiographic grid artifact correction

Area of Science:

• Medical Imaging
• Radiography
• Image Processing

Background:

• Antiscatter grids in x-ray radiography introduce periodic noise artifacts, hindering image quality.
• Traditional spectral domain filtering methods for artifact correction cause blurring and information loss.

Purpose of the Study:

• To develop an effective spatial-domain solution for correcting grid artifacts in x-ray images.
• To improve diagnostic accuracy by preserving image details during artifact removal.

Main Methods:

• Proposed Grid Regression Demodulation (GRD) based on spatial-domain analysis.
• Formulated degridding as a cost functional involving latent image and grid model.
• Developed optimization algorithms for artifact correction and grid pattern estimation.

Main Results:

• GRD effectively removes primary and harmonic grid artifacts from simulated and real x-ray data.
• The method preserves image edges, textures, and patterns, unlike spectral filtering techniques.
• GRD achieved near-perfect correction in ground truth settings with low residual errors and no additional artifacts.

Conclusions:

• GRD outperforms existing transform techniques for digital radiography grid artifact correction.
• The method is self-adaptive and restores the true scale factor of degridded data.
• GRD offers a promising approach for upgrading low-dose medical radiographic imaging technology.