Image Noise Covariance Can be Adjusted by a Noise Weighted Filtered Backprojection Algorithm

Area of Science:

• Medical Imaging
• Image Reconstruction
• Signal Processing

Background:

• The filtered backprojection (FBP) algorithm is a standard method for image reconstruction.
• Image noise covariance in FBP is typically anisotropic, affecting image quality.
• Understanding and controlling noise covariance is crucial for accurate image analysis.

Purpose of the Study:

• To investigate the noise covariance properties of the filtered backprojection (FBP) algorithm.
• To demonstrate that noise weighting can alter and potentially isotropize image noise covariance.
• To develop closed-form expressions for noise variance and covariance in the noise-weighted FBP algorithm.

Main Methods:

• Development of a noise-weighted FBP algorithm.
• Derivation of closed-form expressions for noise variance and covariance.
• Computer simulations using transmission and emission noise models.
• Evaluation of noise covariance with and without noise weighting.

Main Results:

• The noise-weighted FBP algorithm effectively alters image noise covariance.
• Closed-form expressions for noise variance and covariance were successfully derived.
• Simulations confirmed the ability to achieve approximately isotropic noise covariance.
• A parameter in the noise-weighted FBP algorithm emulates Landweber iterations, influencing noise covariance.

Conclusions:

• Noise weighting in the FBP algorithm offers a method to control and isotropize image noise covariance.
• The derived closed-form expressions provide a theoretical basis for noise analysis in FBP.
• Proper selection of the noise-weighted FBP parameter can lead to approximately isotropic noise covariance functions.