Improved tomographic reconstruction of large-scale real-world data by filter optimization.

Area of Science:

• Computational imaging
• Image reconstruction algorithms
• Scientific data processing

Background:

• Large-scale tomographic experiments generate vast datasets, posing computational challenges.
• Limited projection data (low number or low signal-to-noise ratio) often compromises reconstruction quality.
• Existing reconstruction methods offer trade-offs between speed (analytical) and accuracy (iterative).

Purpose of the Study:

• To present and evaluate the SIRT-FBP method for processing large-scale, real-world tomographic data.
• To demonstrate the method's ability to approximate Simultaneous Iterative Reconstruction Technique (SIRT) efficiently.
• To address practical implementation challenges with experimental tomographic data.

Main Methods:

• Application of the SIRT-FBP method, which combines SIRT approximation with Filtered Backprojection (FBP) using precomputed filters.
• Focus on implementation details for large-scale experimental data and common issues.
• Computation of experiment-specific SIRT-FBP filters and their reuse.

Main Results:

• SIRT-FBP filters can be computed efficiently, even for large datasets.
• Precomputed filters are reusable for subsequent experiments.
• SIRT-FBP accurately approximates iterative reconstructions for experimental data.
• Achieves higher accuracy than standard direct analytical methods without increasing computation time.

Conclusions:

• The SIRT-FBP method offers a practical solution for accelerating tomographic reconstruction of large-scale experimental data.
• It provides a balance between the speed of analytical methods and the accuracy of iterative methods.
• Enables more efficient analysis of tomographic datasets, especially those with limited projections.