Re: Depth reconstruction in freeviewed images

[P3D John Bercovitz <bercov@xxxxxxxxxxx>]

I sent a reply in last Thursday morning but it didn't make it out of
the exploder so here's a repost:
 
George Themelis writes:
> I  know this has been discussed in great length in the past, but I think it
> is due for a new round of comments.  My specific question is, how does the
> mismatch of focal lengths affect reconstructed depth?  In particular, what
> effect does the use of significantly longer FLs (generally the case of
> freeviewing or in projection while sitting way back) has on reconstructed
> depth?  I was under the impression that the z-dimension was stretched. 
> That's why portraits look terrible when viewed with longer FLs (projection,
> etc.)  Paul's comments seem to imply that very significant mismatch leads
> to flattening.  Is that because we cannot see the deviation any more? 
> What's is going on?
 
I wonder also.  You are right - the farther back you move, the more the depth 
dimension is stretched.  It's especially easy to see when freeviewing a pair.
Just move closer and farther from the pair and watch the effect occur right
before your very eyes.  This is of course the reconstructed geometry we're
talking about.  Your brain can choose to override whatever it sees based on
prior knowledge of the actual situation.  For a GIF showing the geometry,
turn your web browser to ftp://bobcat.etsu.edu/pub/photo/photo-3d/technical
and look at orthomag.
 
As to Paul's experience, I can only guess that he has gotten so far away
that he's lost the stereo disparity.  See, as you back away, the disparities
get angularly smaller (but paradoxically, this cause more apparent depth, see 
the gif file).  Now if Paul was looking at a low-res pair, the disparity might
become unobservable even sooner since it would be harder for the eyes to pick
the disparities out of the mush.
 
John B


------------------------------