Header banner

<< Previous Thread Llens matching testing, taking advantage of stereopsis Next Thread >>

Subject: Llens matching testing, taking advantage of stereopsis
Date: 2008-02-17 17:46:09
From: DrT (George Themelis)
Here, I gave myself an idea :)

We all know how sensitive our brains are in detecting small differences
between identical pictures, as retinal rivalry or a difference in depth.
Why not use this sensitivity to check focal length matching between
lenses?

The procedure is simple: Use a single (2D) camera. Take pictures of the
same far-away scene with the different lenses. The far away object must
be far away and/or be flat and perpendicular to the camera axis and have
good details, especially at the edges (pictures of the moon will not
work!) Keep good notes!

Then mount the different film chips in a stereo mount and observe in a
stereo viewer. Mount the film chips so that in the left side the window
is at infinity. There are three possibilities:

1- Right side window is also at infinity -> Lenses reasonably well matched

2- Right side window is in front of far object (or object is behind the
window on the right side, or the right eye can see LESS on the right
side) -> Right image is smaller, right lens has shorter FL.

3. Right side window behind the far object (or object is in front of the
stereo window, or eye sees MORE on right side). This situation tends to
confuse some people. There is a non-stereoscopic band on the right side
that cannot be placed properly in space. You can reverse the film chips
and then will have situation #2, which might be easier to see. The
conclusion is that the right image is larger, or the right lens has longer
FL.

I believe that the eyes can detect a 0.3% difference in image size this
way, and that could be the "limit of acceptance". For 35mm film chips,
mounted in 33mm wide RBT mounts, this corresponds to 0.1mm difference in
size. I am going to try this with the Contax/Yashica lenses I have been
"collecting" lately!

George Themelis
Subject: Re: Llens matching testing, taking advantage of stereopsis
Date: 2008-02-18 02:09:11
From: Bill G
Hi Doc

> Here, I gave myself an idea :)
>
> We all know how sensitive our brains are in detecting small differences
> between identical pictures, as retinal rivalry or a difference in depth.
> Why not use this sensitivity to check focal length matching between
> lenses?
>
> The procedure is simple: Use a single (2D) camera. Take pictures of the
> same far-away scene with the different lenses.
Does the camera position change between the two
shots? You did not mention this?



Bill
Subject: Re: Llens matching testing, taking advantage of stereopsis
Date: 2008-02-18 11:26:34
From: DrT (George Themelis)
>> The procedure is simple: Use a single (2D) camera. Take pictures of
>> the
>> same far-away scene with the different lenses.
> Does the camera position change between the two
> shots?

No, it does not. Ideally (if the lenses match perfectly) you get two
identical pictures. But if the lenses do not match, the pictures are not
identical.

You can see 1) difference in exposure (if any), 2) difference in size.
The difference is size manifests itself as a tilted stereo window. The
eyes are very sensitive to this. At least my eyes. (I know some people
cannot see the stereo window very well).

I am actually very excited and ready to try this with my 35mm camera
lenses (I have a lot of lenses to test!) I will write a blog to show you
how this works.

With digital photography it can be easily demonstrated as follows: Take a
2d image. Duplicate it. Enlarge the copy by, say, 1%. Crop the enlarged
copy so the left sides are identical. The right sides are not identical
because the 2nd image is 1% larger and both images are cropped to the same
final size. Try to view this in stereo. The difference is size will show
as a tilted stereo window. With small digital images I would think that
you need 1% difference in size to see what is going on, but with slide
film in a good viewer, with good eyes and good view of the edges, you can
detect must smaller difference. I don't think you need an optical device
(microscope) to carry quantitative measurements (but this is always an
option).

George
Subject: Re: Llens matching testing, taking advantage of stereopsis
Date: 2008-02-18 12:33:02
From: Bill G
Hi doc
>
> No, it does not. Ideally (if the lenses match perfectly) you get two
> identical pictures.
Ok, I assumed this, but wanted to be sure I followed
you.....



>
> You can see 1) difference in exposure (if any), 2) difference in size.
> The difference is size manifests itself as a tilted stereo window.
If the goal is to capture an object to film, to
determine variances in fl of different lenses, it would make more sense
to simply measure the object on film (as discussed in the previous post)
vs. using stereopsis. You can measure with either a microscope or by
scanning the film. Allowing human vision through stereopsis to
determine lens fl variance introduces too many NEW variables into the
process, which must be controlled..... just a few would be:


1) viewer lenses being perfectly matched fl's,


2) the person should have perfectly matched eye fl's,


3) each eye is exactly the same distance from the viewer lens (as
slight variance in distance from the eyepiece will alter the "apparent
fl" of the viewer lens,


4) the viewers eye pupils are perfectly concentric with the viewer
lenses and aligned with infinity points to avoid distortions - which
alters image size. Non concentric viewing will create un equal
distortion, even if small (but remember, the differences you are looking
for are small also)


Also, you are relying on object deviation
ONLY, to determine depth - which is the basis for discovering variance
in lens fl's. However the brain uses many other depth cues other than
object deviation.... in the end, you end up with a witches brew of
variables, vs. photographing the same object by every lens, and simply
measuring the object on film, using magnification to increase the
measuring accuracy.






> I am actually very excited and ready to try this with my 35mm camera
> lenses (I have a lot of lenses to test!) I will write a blog to show you
> how this works.
>
I understand you concept, but I think there
is much easier and MUCH more accurate means to determine what you are
chasing.





> With digital photography it can be easily demonstrated as follows: Take a
> 2d image. Duplicate it. Enlarge the copy by, say, 1%. Crop the enlarged
> copy so the left sides are identical. The right sides are not identical
> because the 2nd image is 1% larger and both images are cropped to the same
> final size. Try to view this in stereo. The difference is size will show
> as a tilted stereo window. With small digital images I would think that
> you need 1% difference in size to see what is going on, but with slide
> film in a good viewer, with good eyes and good view of the edges, you can
> detect must smaller difference. I don't think you need an optical device
> (microscope) to carry quantitative measurements (but this is always an
> option).
>
Or, if you already have the
film, just scan it 4000+ ppi, then measure the distance of the object in
PS? There is way too many variables (which most of them you can't
nail down) to rely on stereopsis to determine variance in fl between
lenses In addition, the group was trying to compare many lenses at once
to best match their fl's.... Again, the measurement of an object
provides a non-subjective basis for comparing the variance in lens
fl's. Why over complicate the procedure?
Subject: Re: Llens matching testing, taking advantage of stereopsis
Date: 2008-02-18 13:28:48
From: Ken Strauss
Using a scanner to measure introduces another variation depending on how you
scan. You should always position the originals in the same place on the
scanner and measure perpendicular to the axis of movement of the scanning
mechanism.

> -----Original Message-----
> From: MF3D-group@yahoogroups.com [mailto:MF3D-group@yahoogroups.com] On
> Behalf Of Bill G
> Sent: Monday, February 18, 2008 1:33 PM
> To: MF3D-group@yahoogroups.com
> Subject: Re: [MF3D-group] Llens matching testing, taking advantage of
> stereopsis
>
> Or, if you already have the
> film, just scan it 4000+ ppi, then measure the distance of the object in
> PS? There is way too many variables (which most of them you can't
> nail down) to rely on stereopsis to determine variance in fl between
> lenses In addition, the group was trying to compare many lenses at once
> to best match their fl's.... Again, the measurement of an object
> provides a non-subjective basis for comparing the variance in lens
> fl's. Why over complicate the procedure?
Subject: Re: Llens matching testing, taking advantage of stereopsis
Date: 2008-02-18 14:20:08
From: DrT (George Themelis)
> Allowing human vision through stereopsis to
> determine lens fl variance introduces too many NEW variables into the
> process, which must be controlled..... just a few would be:
> 1) viewer lenses being perfectly matched fl's,
> 2) the person should have perfectly matched eye fl's,
> 3) each eye is exactly the same distance from the viewer lens (
> 4) the viewers eye pupils are perfectly concentric with the viewer

I know this is going beyond MF3d photography but I just want to say that
these variables do not affect the result of my visual testing.

The visual testing I described is similar to the famous "fly" or similar
stereopsis tests given by doctors to test stereopsis. The differences are
built into the image. There is only one question: Can you see it, or
not? When the doctor asks "Is object A closer than object B?" there is
only one correct answer. Your eyes, glasses, filters, distance from the
image, etc., have no effect on the answer, only your ability to see the
answer. And if you give the wrong answer, the conclusion is that you are
guessing, or cannot see it.

Of course, I realize that some people can see it and some cannot. Since I
am using my own lenses for my own enjoyment, I guess the only question
that maters to me is "can I see it?". (BTW, I pass the stereopsis tests
with flying colors :))

Bill, you previously stated that a mismatch in taking lenses and a
mismatch in the viewing lenses are the same thing. I used to say that too
(or I could not really express the difference between the two situations)
until this was discussed in photo-3d and I had some new insights. I
forget that these insights were, but one thing is clear now: When you
have a mismatch in taking lenses and you put the images in a fixed frame
of reference (= stereo mount), you can see the mismatch of the taking
lenses with respect to this fixed frame. A mismatch in the viewing lenses
will not change the relationship between the images and the fixed frame
because it affects both the same way.

> Again, the measurement of an object
> provides a non-subjective basis for comparing the variance in lens
> fl's. Why over complicate the procedure?

I find the concept of using our own visual system and enchased stereopsis
fascinating (finally, we use the skills we have developed by looking in
stereo pictures, for something practically) and simpler (no
measurements!), at least for my own needs. But for a group project, I
agree that it is better to follow the quantitative method.

An equivalent situation from a different area of stereoscopic research:
When comparing projection screen material to see which one is brighter or
has better extinction, you can use a light meter and measure the reflected
light from each screen under various conditions, or you can just put two
screens next to each other and look at them to see which one is brighter.
Comparing the screens visually in pairs, is a surprisingly accurate method
to compare screens and you have the satisfaction that you are actually
seeing the differences! (plus you can see some aspects right away, like
uniformity of the light, fall off in the edges, etc, that are a bit more
complicated to measure - but I agree that anything that can be seen can
also be measured).

I am signing off now :) I will try the visual testing system with my
lenses and, if I have anything more to say, I will take it up in photo-3d.

George
Subject: Re: Llens matching testing, taking advantage of stereopsis
Date: 2008-02-18 19:54:05
From: Bill G
Hi Doc


>> Allowing human vision through stereopsis to
>> determine lens fl variance introduces too many NEW variables into the
>> process, which must be controlled..... just a few would be:
>> 1) viewer lenses being perfectly matched fl's,
>> 2) the person should have perfectly matched eye fl's,
>> 3) each eye is exactly the same distance from the viewer lens (
>> 4) the viewers eye pupils are perfectly concentric with the viewer
>>
>
> I know this is going beyond MF3d photography but I just want to say that
> these variables do not affect the result of my visual testing.
>

Not sure how you can take this
position. Think of it in reverse.... If you have two identical sized
imaged objects, and you have viewer lenses which are NOT matched
magnifications, the images displayed on your retinas will be different
sizes (assuming your eye fl's are identical) The same is true of all
the other variables I listed above....and several others I did not
list. It's the retinal image that matters, and there is several links
in the optical chain which produces the retinal image. All links in
the chain are equally significant....sort of the "weak link in the
chain" principle. You seem to be singling out ONE of the variables
and disregarding all the other variables that produce the retinal image.





> The visual testing I described is similar to the famous "fly" or similar
> stereopsis tests given by doctors to test stereopsis. The differences are
> built into the image. There is only one question: Can you see it, or
> not? When the doctor asks "Is object A closer than object B?" there is
> only one correct answer. Your eyes, glasses, filters, distance from the
> image, etc., have no effect on the answer,

Just one example here....if your eye fl's
errors are significant enough, it can alter the size of the two
images. This can hamper your brains ability to comprehend which object
is closer. The same is true of the other variables.... I am not an
optician, but I can assure you, the set-up they use is compensating for
any of these errors which otherwise would ruin the test.






> only your ability to see the answer. And if you give the wrong answer, the conclusion is that you are guessing, or cannot see it.
>
Of course using very high end testing equipment
in the medical field whereas the design engineers have already accounted
for all these variables...... yes, I can see this test being quite
accurate.....but I think its a stretch to think a $29 Chinese MF viewer
with no QC, no correction for refractive errors, no IPD adjust, will
provide equal results.






> Bill, you previously stated that a mismatch in taking lenses and a
> mismatch in the viewing lenses are the same thing. I used to say that too
> (or I could not really express the difference between the two situations)
> until this was discussed in photo-3d and I had some new insights. I
> forget that these insights were, but one thing is clear now: When you
> have a mismatch in taking lenses and you put the images in a fixed frame
> of reference (= stereo mount), you can see the mismatch of the taking
> lenses with respect to this fixed frame. A mismatch in the viewing lenses
> will not change the relationship between the images and the fixed frame
> because it affects both the same way.
>

I agree with your reference frame in
theory.... however, the brain will overcome a lot of these variances in
retinal images..... this is a result of the brains ability to be
influenced by other cues, not just object deviation and frame reference.


If you want to run an experiment, to
witness how one "weak link in the optical chain" can radically alter the
brains ability to interpret retinal images.....try this....


View two grids patterns, such as 1/4" grid
paper. Be sure you have a jig whereas the lens centers are concentric
with your IPD. Align each grid so you can perfectly fuse the grid
pattern. Use a reference to be assured your axis's are true to each
other, as well as your lens axis. Now, when you swivel your eyes to
the left, you will notice some lens distortion. But the distortion
should be equal in both retinal images as the lenses are both spherical
- and you started with your eye pupils concentric with the viewer
lenses. The image perceived will be in 2d as there is no object
deviation and you are viewing objects (lines) with no variance in their
Z axis. Unlike viewing images of spherical shapes, whereas the brain
has some Z axis data to work with.


Now, move the left viewer lens with its
associated grid wider than your IPD, maybe 4-5mm. The grid must remain
true to the lens. Now view the same left edge. You have two
different distortion patterns delivered to the retina. Why? Each eye
pupil axis is intersecting the viewer lenses in different areas, as we
lost the concentric set-up we started with. The right eye will be
viewing through the nasal portion of the viewer lens, while the left eye
will be viewing on-axis through the left lens, hence two different
distortion patterns. Now the grid pattern appears in a wavy pattern
on the Z-axis, very similar to swells in the ocean. This is the brains
3-d interpretation of object deviation of identical 2d images. The
brain has no good reference from previous / historical viewing. It's a
freaky view.


Now run the same experiment using two good
photo stereo images. No Z axis distortion! Why? Because the brains
knowledge of how scenes are compiled will correct the Z-axis distortion
and deliver a good (or at least decent) stereo image, unlike the wavy
Z-axis graph paper. However, this continuous error compensation puts a
heavy load on the brain... eventually leading to discomfort - which
was somewhat the basis of this thread, (i.e. what the importance of
matching taking lenses)


This experiment demonstrates 3 points..... 1)
how critical all the stereo variables are to the stereo process (from
taking to viewing).... and, 2) it only takes one variable such as
unequal distortion as explained above, to hamper our ability to use
stereopsis as a measuring or diagnostic tool, and, 3) the brains
unique ability to correct many of the errors we introduce into the
stereo chain, albeit, often at a price.
Subject: Re: Lens matching
Date: 2008-02-20 01:46:33
From: depthcam
Oh, shucks... I was only away for a few days and I come back and it
looks like I started a war !!! :-(((

OK Doc,

You seem to assume that because you were fairly lucky with your lens
matching, it means there is no such thing as unmatched pairs of single
focal length lenses. Let me tell you about my own experience:

Back in 1980, I started out using fifties stereo cameras and, among
those I had a Verascope F40 that produced film chips of different
image size. Then I had - of all things - a "modern" Nimslo where I
used the two outermost lenses for regular slide pairs and - surprise:
mismatched lenses !

In 1982, I had a set of Minolta X-500 cameras spliced and made into a
stereo camera. I was in fact the first person to have this done with
an X-500 (by a Frenchman named Franck Loriot - before Ton Pennings did
a similar job for David Burder).

Because of my past problems with mismatched lens pairs, I contacted
Minolta Canada and asked them if they could supply me with matched
sets of lenses in 35mm, 50mm and 100mm focal lengths. They answered
that the matching could only be done at the main manufacturing plant
in Japan. So I had to special order them and it took two months to
get to me. Interesting that Minolta Canada could not guarantee a
perfect match from their stock lenses when you imply than any off-the-
shelf set of lenses picked at random will do the trick.

Back in the early nineties, I worked on a project for an MF 6x7 stereo
camera with Hermann Miller in Germany. The camera was outfitted with
Schneider Super-Angulon lenses. We had Schneider match the lens pairs
and each lens came with a matching certificate that stated the
effective focal length. In my case, each lens was 47.53mm

I also worked in audio visuals in the eighties and all our projector
lenses came with effective focal length certificates so the images
would match on the screen.

When spending large amounts of money on MF lenses, it's not a bad idea
to make sure they will be an acceptable match. Sure, you may get
lucky, but you may not...
Subject: Re: Lens matching
Date: 2008-02-20 01:56:59
From: DrT (George Themelis)
----- Original Message -----
From: "depthcam" <depthcam@yahoo.ca>


> Oh, shucks... I was only away for a few days and I come back and it
> looks like I started a war !!! :-(((

More like a discussion...

Do you have a name? It is a good idea to sign your emails....

Thank you,

George Themelis