Automatic regularization parameter selection by generalized cross-validation for total variational Poisson noise...

Area of Science:

• Medical Imaging
• Computational Science

Background:

• Photon-counted imaging is crucial for low-light conditions.
• Accurate image reconstruction is vital for diagnostic quality.
• Existing methods require manual parameter tuning, limiting efficiency.

Purpose of the Study:

• To develop an advanced image reconstruction algorithm for photon-counted images.
• To automate the selection of regularization parameters within the reconstruction process.

Main Methods:

• An alternating minimization algorithm was developed.
• Generalized cross-validation was employed for automatic parameter selection.
• The algorithm updates parameters iteratively during reconstruction.

Main Results:

• The proposed algorithm demonstrated superior performance compared to existing methods.
• Automatic parameter selection improved reconstruction accuracy and efficiency.
• Outperformed Maximum Likelihood Expectation Maximization (MLEM) and primal-dual methods.

Conclusions:

• The novel alternating minimization algorithm offers a significant advancement in photon-counted image reconstruction.
• Automated regularization parameter selection enhances robustness and accuracy.
• This method provides a more efficient and effective solution for low-light imaging applications.