Automatic image-domain Moiré artifact reduction method in grating-based x-ray interferometry imaging.

Area of Science:

• Medical Imaging
• Computational Imaging
• X-ray Interferometry

Background:

• Grating-based X-ray interferometry is a powerful imaging technique.
• System instabilities can induce Moiré artifacts, degrading image quality.
• Current methods focus on signal retrieval, which can be complex.

Purpose of the Study:

• To propose a novel Convolutional Neural Network (CNN) based method for Moiré artifact removal.
• To improve image quality in grating-based X-ray interferometry.
• To offer an alternative to traditional signal retrieval optimization algorithms.

Main Methods:

• A CNN model was trained using a synthesized dataset.
• The dataset included natural images and experimentally acquired Moiré artifact-only images.
• A novel approach was developed to generate diverse Moiré artifact images from experimental data.

Main Results:

• The developed CNN method effectively removed Moiré image artifacts.
• Both numerical and experimental results validated the CNN's performance.
• Significant improvement in image quality was observed in grating-based X-ray interferometry.

Conclusions:

• The CNN-based image post-processing is an effective strategy for Moiré artifact reduction.
• This method offers a practical solution for enhancing X-ray interferometry imaging.
• The approach holds potential for improving various X-ray imaging applications affected by artifacts.