Regridding reconstruction algorithm for real-time tomographic imaging.

Area of Science:

• Synchrotron radiation science
• Advanced imaging techniques
• Computational microscopy

Background:

• Third-generation synchrotron sources enable sub-second temporal-resolution tomographic microscopy.
• High data rates (near 10 GB/s) create significant data handling and post-processing bottlenecks.
• Existing reconstruction algorithms hinder the full exploitation of ultrafast acquisition speeds.

Purpose of the Study:

• To present the fast reconstruction algorithm gridrec, optimized for conventional CPU technology.
• To demonstrate gridrec as a viable alternative to standard filtered back-projection methods.
• To address the data processing bottleneck in high-speed tomographic microscopy.

Main Methods:

• Implementation of the gridrec algorithm, based on Fourier transform methods.
• Utilizing a regridding procedure for resampling Fourier space from polar to Cartesian coordinates.
• Performance and accuracy comparison against standard filtered back-projection routines.

Main Results:

• Gridrec offers a valuable alternative to filtered back-projection with negligible accuracy degradation.
• The algorithm achieves up to a 20-fold performance increase.
• Gridrec's signal-to-noise ratio superiority in well-sampled tomographic problems was demonstrated.

Conclusions:

• Gridrec overcomes data processing bottlenecks in high-speed tomographic microscopy.
• The algorithm enables real-time monitoring of sub-second acquisition processes.
• Gridrec facilitates the full potential of ultrafast synchrotron imaging.