[photo-3d] Re: Base Calculator, Deering/Bercovitz math

[Abram Klooswyk <abram.klooswyk@xxxxxx>]

Tom Deering again discussing with Dr.T.! Keep your fingers
crossed. :-)

Joseph Valvo, who started this branching thread, recommends 
Tom Deering's math page, and thanks John Bercovitz for his 
"early work".  (17 Sep 2000, Stereo Demystified (originally 
base calculator))

So it's time for some relativization :-) 
At the ISU Congress in Arlington/Washington I have met a 
member of the Potomac Society of Stereo Photographers who 
was a great expert on history. However, another American once 
seems to have said "all history is bunk" (Henry Ford).

So it seems that several Americans believe that Oliver Wendell 
Holmes invented the stereoscope :-). And it seems that several 
others believe that John Bercovitz invented stereobase math. 
However, almost _all_ of that math goes back to formulas and 
principles from early in this (still the 20th) century, some 
root even in the 19th century. Some day I might give more 
details, but calling John's work "early" is about the same 
as calling Washington Cathedral a medieval building. 

But now for the flaws. 
The basic idea behind the MAOFD theory is that the far point 
distance is one of the determinants of the stereo base. This 
indeed is a very old idea with has done much harm when used 
for scenes (at _any_ distance) with a limited depth extension. 
This idea resembles a Procrustian bed. 
Try for example a scene with kids playing close just before 
a wall. They come out stretched to infinity. 

Or read my pun "MAOFD great for mono wide-screen macro!" at 
www.pauck.de/archive/mailinglist/photo-3d/mhonarc/msg36987.html
(I used a 1:1 macro setup with a full frame camera on slide 
bar for a woodcarving with 10 mm depth, the MAOFD formula 
arrived at a base of 36.3 mm, just enough to have NO common 
image content in both pictures, but a great 24 x 72 mm MONO 
panorama).

So for macro with limited depth the MAOFD formula and graphs 
give just absurd and useless results.

Bob Maxey 15 Sep 2000: 
God's Natural Interocular Separation "works well for 90 
percent of subjects. I have seen a few of these and they all 
seem to disagree with the other formulas I have seen."

Indeed. When the MAOFD formula is applied to all those 
pictures made with standard stereocamera's, for the majority 
of the pictures the formula suggests a different base than the 
camera uses. 
This is absurd of course. The formula adepts say something 
like: "Yes, but it is on _maximum_ deviation, not on 
_desirable_ deviation". 
Great Formula!, it gives you a figure you shouldn't use :-).

On his page Tom Deering indicates when the 1/30 rule doesn't 
apply, among them "hyperstereo". 
However, in hypers mostly distant or very distant objects are 
included. Then the 1/30 rule works very well, the MAOFD 
arrives at the _same_ base only after putting in all those 
variables. 

When the distance of the far point is very large, 1 divided by 
that far point distance is very small. Therefore it doesn't 
have very much effect on the base in the MAOFD math. 
But the 1/30 rule doesn't use the far point at all, so gets 
there quicker.

So MAOFD theory is not needed for hypers, not usable at normal 
distances and absurd for subjects with limited depth, as often 
in macro. 
When use it at all? It is great when you deliberately 
want to increase depth, in cases you don't have to fear for 
accusation of bad taste or kitsch shooting. Sir David Brewster 
said: "To add an artificial relief is but a trick which may 
startle the vulgar, but cannot gratify the lover of what is 
true in nature and in art." For that trick the MAOFD formula
is great.

Another (small) point on Tom Deering's site is the 
"1/f rule of thumb". 
>> You can use this rule in landscape situations because it 
takes into consideration the focal length "f" of the camera's 
lenses. This becomes the 1/35 rule for 5-perf cameras, the 
1/50 rule for 35mm cameras, and the 1/80 rule for medium 
format. While it still only works for landscape photos, using 
the 1/f rule will work with larger and smaller formats, and 
longer and shorter lenses.<<

Note that Tom doesn't recommend this rule. All the better,
for it is the Mother of Illogical Rules. 

For MF, and when viewer focal length = camera focal length =
80 mm, the standard stereo viewing space principle says: 
Standard maximum deviation is 80 / 30 = 2.7 mm.
But the 1/f rule gives a stereo base of only 25 mm at 2 meter 
(7 feet) distance from the camera, which results in a near
point deviation of only 1 (one) mm, less than the standard
for the Realist!  Need I say more? :-)
The only time this rule works is when viewing focal length = 
camera focal length = 30 mm, for then it is the 1/30 rule :-)

A rule that ignores viewing is bound to be useless.
Finally, do I really believe in this 2 degrees viewing space?
As a guideline, yes. 

Abram Klooswyk