I have been trying to perform a blind deconvolution on 3D image stacks I have taken using a serial sectioning WF fluorescent imaging technique I’ve been developing.
One of the main problems I have is that the imaging I perform is not a straight forward wide-field fluorescence but has a sectioning component. In brief, a fluorescent-stained sample is embedded in a hard resin block; an image is taken of the block surface, then a thin (1um) section is automatically cut and a another image take etc. what you end up with is a comet tail like artefact in one z direction only.
The image above shows the orthogonal views of cell clusters, with the comet tail artefact visible.
The result is an image stack which have a comet tail like artefact in the xz and yx planes (see image above). I am quite a novice at deconvolution but I believe that this asymmetry the source of my problems with techniques I have tried so far.
I have spent some time trying various deconvolution plugins (mostly based on RL blind deconvolution) and I implementing a variation on a nearest neighbour (one direction only) deblurring method but I have not had much luck: The nearest neighbour method had issues with noise enhancement, and the blind RL is not finding a PSF that looks vaguely sensible. I recently tried the EpiDEMIC plugin on icy and the result was a bit more promising but again the PSF was symmetric and resulted in an odd looking image where the cell I was looking at was split in two and mirror imaged in the z plane. Having used the Forums for that plugin again the issue seems to be the asymmetry of the PSF I need.
Due to technical difficulties I cannot image sub-resolution beads to get a PSF and so am looking for a blind solution. Does anyone have any ideas as to how I could best approach this?