Three-Dimensional Reconstruction of Noisy Images from Partially Confocal Scanning Microscopes

In three-dimensional microscopy the axial resolution is about an order of magnitude worst than the lateral resolution. Two common, solutions are either to process the recorded image with a deconvolution algorithm or to record the image with an instrument with better depth discrimination, such as the confocal scanning microscope (CSM) where the confocal aperture, a pinhole in front of a light detector, reduces the out-of-focus contributions. The improved axial resolution obtained by either method is still worst than the lateral resolution by a factor of three. In this work we investigate the combination of confocal scanning microscopy with posterior processing to achieve equal lateral and axial resolution. A deterministic and a stochastic algorithm are applied to synthetic images supporting that it is possible to achieve the same resolution laterally and axially, and that the limiting factor is noise. Both algorithms are nonlinear and thus potentially capable of recovering some of the information lost in the image formation process (superresolution). In this work we also investigate the superresolving capabilities of the two algorithms. Strictly confocal microscopy relies on an infinitesimally small confocal aperture. This is neither possible nor desirable since such a small aperture will collect little or no light. A finite size aperture results in partially confocal imaging. Smaller aperture produce better depth discrimination, but reduced signal-to-noise ratio (SNR); wide apertures have poor depth discrimination but better SNR. The trade-off is investigated and no clear optimum size is found. Simulations are presented indicating that the selection of the size of the confocal aperture must take into account what the reconstruction algorithm can achieve with the resulting combination of prior depth discrimination and SNR.