Adaptive frequency-domain regularization for sparse-data tomography

摘要

A novel reconstruction technique, called Wiener Filtered Reconstruction Technique (WIRT), for sparse-data tomographic imaging is introduced in this article. This six-step method applies a spatially-varying constrained least-squares filter combined with a regularization method based on total variation. The WIRT reconstruction is implemented in the frequency domain, where the information based on measurements and regularization can be treated separately. The algorithm applies regularization selectively in the frequency regions where the frequency component values cannot be defined by the measurements. This leads to computational benefits when compared to conventional iterative reconstruction methods such as algebraic reconstruction technique (ART). Both qualitative and quantitative comparisons against state-of-the-art methods suggest that WIRT is a promising reconstruction algorithm for sparse-data imaging regimes, especially with higher noise levels.