Alfred Molon wrote:
In article , Floyd L. Davidson says...

But probably the most useful would be Richardson-Lucy
Deconvolutional sharpening. Using just standard default
filters (Gaussian and perhaps exponantial) for the point
spread function should work better than other sharpen
tools, but it would also be possible to develop a very
accurate point spread function for any given specific
lens (think of the Hubble Telescope), and that would be
very significantly better than other methods.

If the softness is caused by a particular camera-lens combination, is
there a way to determine this point spread function with for instance
some calibration steps, and then apply the right sharpening/
deconvolution?

Any good optical engineer could work it out, given the
right equipment and a fairly fat check.

That's not a likely route for any but the most serious and
well healed.

And in fact using a Gaussian spread will work rather well
for a slight defocus problem, for diffraction, and for
"haze" from atmosphere.

I don't know for sure, but seem to remember that there
is software available that will measure motion blur and
might be able to produce a custom point spread function
to counter it.

I'm even wondering if this point spread function could be determined
from an image (without any previous camera/lens calibration), so that
the image processing software can choose the right sharpening algorithm
and parameters.

Some aspects might be. To really do it right though
you'd need to have the lens on an optical bench, or have
some very specific design specifications.

But sharpening by inspection using a default Gaussian
blur for the point spread function does do a fairly good
job of sharpening. I'm sure you can use Google to find
software that will let you try it. GIMP can do it with
a GMIC plugin.

--
Floyd L. Davidson http://www.apaflo.com/
Ukpeagvik (Barrow, Alaska)