What is the computation time for deconvolving an image?

• Microscope type: WF microscopes require more iterations than confocal or 2-photon microscopes.
• Object type: sparse objects can be restored more effectively than dense objects. The more gain is possible, the more iterations are needed, even if the iterations themselves become also more effective ('bigger steps').
• Noise: low noise makes a large gain possible: more iterations are needed.
• Algorithm: Our ICTM (Iterative Constrained Tikhonov-Miller) iterations take per iteration less time than our MLE (Max Likelihood Estimation) algorithm.
• Hardware: faster and more processors speed up things, insufficient memory is problematic: processing speed then depends on disk IO performance.
  • Example: Restoring a moderately noisy confocal 256x256x64 image, including starting the software, loading the image, generation of a PSF and a MLE run take altogether 4:23 minutes on a SGI Octane with 2xR10000@225MHz