[photo-3d] Re: Algorithm Wanted

[Bruce Springsteen <bsspringsteen@xxxxxxxxx>]

An excellent variety of responses so far, reflecting the talents and
temperaments of the writers.

I missed Olivier Cahen's fine response earlier, due to the subject line. 
His method for trimming and aligning typical separated prints sounds
similar to the one I used in my first stereo attempts (based on the second
mounting method given in the second chapter of Ferwerda), including
Olivier's use of the View Magic viewer.

This method assumes that the prints as received are already close to being
in correct trim, so doesn't really serve as the fool-proof procedure
wanted here, but it is close enough in many circumstances.  Honestly, I'm
not interested in "seeing vectors", but any method of aligning stereo
prints requires the discovery of the displacement direction, whether we
use the term "vector" or not.  I'm on record here as not heeding
distinctions between theoretical and practical inquiries; I consider them
complementary aspects of all problems, neither of which can be ignored in
the long run.  So I ask about both together.

[I'm happy to see one more person describing what an excellent tool the
View Magic, or other mirrored print viewer, can be in working with prints,
quickly and dynamically resolving alignment/cropping questions (shifting
and rotating prints, masking for window with paper strips, all while
viewing).  Why printmakers work without availing themselves of this tool,
even if they intend to mount in Holmes format, is a mystery to me.]

Abram Klooswyk admits to enjoying my "seemingly useless" questions, so I
assume they are at least useful as conversation pieces.  They never are
easy to simply dismiss, are they?  The fact that knowledgeable stereo
people have evidently already considered this problem long before me
suggests that it's perhaps not a complete waste of time! ;-)

I pose these problems without providing answers because I have half-formed
ideas that I cannot make complete on my own.   I launch P3D "fishing
expeditions" hoping that those older or cleverer than me will clear the
blockage, and save me from clumsily reconstructing 30 years of existing
work.  It generally succeeds, so I continue to do it!  What a sensible use
of this mailinglist!  My questions are usually prompted by some practical
problem that has set my teeth grinding.

I like Abram's vectors posing as wooden arrows, but his gruesome X-ray
scenario seems not so pretty as my star-field reverie.  I do wonder what
he would devise as original solutions, if his historical knowledge weren't
so complete and readily accessed. ;-)  In any case, the superposing of two
homologues had occured to me, as well as the rotation around them, but I
was unsure what to do next.  I suspect there may be a way to perform this
operation on prints without overlaying, but I will have to experiment. 
Perhaps something as simple as laying tracing paper over each print
successively and marking x's and drawing lines.  Gabriel is right that you
can overlay prints on a strong light.   I find that more difficult that he
suggests, and I have tried it before - it's hard to see those points in a
photograph through the paper grain, which tends to diffuse details.  This
would work better with stereo line drawings.

Thanks folks.  You have all given me some more to think about.  Any ideas
from new people?  From inveterate lurkers?  Do you have any comments on
each others' methods?  Or has the question been exhausted?  There seem to
be many related but significantly different ways to approach this basic
stereo problem.

Bruce



__________________________________________________
Do You Yahoo!?
Yahoo! Mail - Free email you can access from anywhere!
http://mail.yahoo.com/