Hullo All,
After a lot of experiments and reading many learned papers...much of
which flew way overhead...I am becoming easy to convince that there
may not be a general (and relatively easy to implement) solution to
the problem posed.
I have to say that there are still many aspects of the problem that I
find inscrutable. A few things don't make sense to me.
To recap, I can carefully align the full blurred and sharp test images
(working with a test set of the trees/woods provided by JohnG), find
the exact same pixel in each and extract a pair of sub-images
(256x256), and then derive a PSF which can be used to generate an
(almost) exact replication of the sharp from the blur. All the theory
stuff works perfectly up to this point.
If I then extract a larger (512x512) image from the aligned blur and
apply the PSF to the 256x256 tiles making up this larger test image,
the resulting derived 'sharp' is poor (more like dreadful!).
Some of the stuff I have read about this implies that because real
images are of the 3D world the varying depth of each pixel will mean
that there isn't a constant PSF across the image and that additional
(very inscrutable) stuff has to be done to arrive a reasonable
correction.
If this were true for the case I have been spending most of my time
on, then the result recreating the 256x256 sharp image might be a bit
of fluke, but that result can be reproduced for a few different
256x256 sub-images.
What is more confusing is that if I take a 512x512 blurred image which
includes the 256x256 image used to derive the PSF in one corner (maybe
not exactly, but to within 1 or 2 pixels), applying the PSF to each
256x256 tile produces a rubbish result for each tile!
Why is so (as Julius Sumner Miller would have growled)?
Blowed if I know. Either the derived PSF can only be applied to the
exact same image it was derived from; I am doing things incorrectly;
something else, or all of the above!

Cheers,
Terry


-- 
You received this message because you are subscribed to the Google Groups 
"Hugin and other free panoramic software" group.
A list of frequently asked questions is available at: 
http://wiki.panotools.org/Hugin_FAQ
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to 
[email protected]
For more options, visit this group at http://groups.google.com/group/hugin-ptx

Reply via email to