A Fourier-domain algorithm for total-variation regularized phase retrieval in differential X-ray phase contrast...

Area of Science:

• Medical Imaging
• Computational Physics
• Image Processing

Background:

• Differential X-ray phase contrast imaging requires a 1D integration step for phase retrieval.
• Standard integration methods are susceptible to noise, causing blurring and streak artifacts.
• Existing regularization techniques often struggle with integration-specific noise characteristics.

Purpose of the Study:

• To develop a novel, regularized integration method for differential X-ray phase contrast imaging.
• To address noise-induced artifacts and blurring in phase retrieval.
• To improve the accuracy and quality of reconstructed images.

Main Methods:

• A Fourier-domain algorithm was developed for regularized integration.
• The method incorporates frequency-dependent noise amplification and 2D data.
• Minimization of total variation orthogonal to the integration direction was employed.

Main Results:

• The proposed method significantly reduced artifacts in both simulated and experimental data.
• No increase in image blurring was observed compared to standard methods.
• Superior performance was demonstrated over conventional integration and image-domain regularization techniques.

Conclusions:

• The novel regularized integration method offers superior artifact reduction for differential X-ray phase contrast imaging.
• This approach enhances image quality by mitigating noise effects without compromising resolution.
• The Fourier-domain technique provides a robust solution for accurate phase retrieval.