[photo-3d] Re: Lenticular depth resolution (was: Camera separation ...)

[Abram Klooswyk <abram.klooswyk@xxxxxx>]

Dear Olivier,

I understand that you don't have Okoshi's book at hand now. 

>the book is entirely written as a theory of holograms. 
(Olivier Cahen, 09 Jul 2000, in reply to my post of 
08 Jul 2000, Lenticular depth resolution (was: Camera 
separation ...)) )

This is only partly true. In his about 400 page book one 
whole chapter is on holography, 109 pages, three other 
chapters have sections on holography, in total another 36 
pages, so in fact little more than one third of the book is 
on holography. Of the remaining, about 129 pages are mainly 
on lenticular systems, but again that topic comes about in 
various other chapters.

I agree of course that binocular resolution is superior to 
monocular resolution, and depth resolution is better than 
follows from calculations using only monocular resolution. 
As you might recall, I have disputed Pizon's calculations 
(in the Bulletin du Stéréo-Club Français) exactly on this 
point, many years ago (before the internet even existed:-)).

However, I suppose you agree that a significant higher 
monocular resolution also gives a higher depth resolution.

>The resolution of the image on the film is not an important 
>parameter of the final depth resolution, since the results 
>were almost the same on images that I took so that it was 
>fuzzy on the film.

This must have been just a little fuzziness. Is has been 
demonstrated in controlled experiments that stereoscopic 
acuity is lower when images are degraded with haze filters, or 
by reducing contrast. (Interestingly the effect of monocular 
degradation is worse than that of binocular degradation.)

You mention stereo acuity figures of your family, worth a 
congratulation :-). However, the method:
> showing (...) a wood bar with two lines of nails, so one 
> could tell which of them was on each line
seems a little dubious.

It is not enough to ask which depth interval people can see, 
when there is no control. In controlled experiments people 
must get a chance to fail :-), that is to discriminate (or 
not) between zero depth and a little depth, and then perform 
better than chance. 
I believe figures will be somewhat worse when the test is done 
in this way. Especially in the case of family members it is 
essential too that the experimenter doesn't know the right 
answers at the time of the test (double blind testing).

The same applies to the test with "the stereo slides of this 
bar". Not double blind testing means that the results always 
will be biased towards the desired result (everyone wants to 
be good at such tests :-)).

This does not imply that the depth resolution figures you 
estimate are necessarily far off, my (rather conservative) 
estimate is also not based on real testing, so large 
differences between estimates are to be expected.

Moreover, the test with nails in a bar will probably give 
higher figures than can be expected in more conventional
three-
dimensional surroundings. It is a well known fact that such 
test objects tend to give the best figures, not attainable in 
most circumstances. My estimate was meant to apply to average 
slides, in order not to overestimate the difference with 
lenticulars.

For the point of my posting was not the exact amount of 
possible depth resolution, but the fact that there is an 
essential difference between twin-view images (stereoslides or 
stereoviews) and lenticulars. In the former 2D resolution is 
identical with number of resolvable line pairs per mm, but not 
so in lenticulars, where the lenslet diameter is one of the 
most important factors (I will not quote all other of Okoshi's 
optical considerations).

Since film resolution is significant better than the
coarseness 
of lenticular sheets, depth resolution in lenticulars will 
never be as good as in twin-view stereopictures.

(One of my unfinished projects in stereoscopy :-( is to design 
an experiment to prove or disprove this statement.)

Abram Klooswyk