Hi:
I always find it useful to experiment with the restoration tool first (which applies different kinds of deconvolution algorithms which are VERY fast) in order to get an idea what PSF to use with the deconvolution tool.
The easiest way to start is to frame a small preview of an area of interest of the main image, and use the restoration tool on the preview to experiment. I always start with deconvolution kernals smaller than what is really needed (small standard deviations for the parametric PSF) and with a shape of 2.0. If you start "over deconvolving" (specifying a spread which is much more than the actual spread in your image) you will get bad results from the get go, so I always start smaller than what actually is probably needed and creep up until artifacts start to appear. A good deal of improvement can be made before artifacting. Once you get a sense for the size and shape you can use the similar parameters in the decon tool.
Note, varying the shape often requires variation of the standard deviation parameter, i.e. these cannot be "optimised" independently of one another...
For the restoration tool the measures of error ranges I find useful are 5 x 10e-3 to 5 x 10e-5 although this depends on the noise in your image.
Please note, for decon and restoration to work properly you should apply them on linear images before any intensity scaling for viewability. Ideally it also should be applied to images without oversaturation of brightness (clipping). I also find converting into 64 bit provides better results (higher precision).
When you finally go to the deconvoution tool, regularisation is very useful.
Just my 2 cents.
PS if you decide to measure the PSf with the dynamic PSF generation tool, one way to massage the generated synthetic PSF to avoid over deconvolution is to scale it slightly smaller (resize it to 100x then scale by 90% 80% 70% 60% etc. and shrink back down to 1/100).
cheers
Colin
