What is a good amount of stereo separation?

A forum to ask questions, post setups, and generally discuss anything having to do with photomacrography and photomicroscopy.

Moderators: rjlittlefield, ChrisR, Chris S., Pau

Smokedaddy
Posts: 1937
Joined: Sat Oct 07, 2006 10:16 am
Location: Bigfork, Montana
Contact:

Post by Smokedaddy »

Thanks

BugEZ
Posts: 850
Joined: Sat Mar 26, 2011 7:15 pm
Location: Loves Park Illinois

Post by BugEZ »

Rik wrote from his updated documentation in Zerene:
For example, if you want a viewing angle that is 4 degrees off-axis, with a stack whose frame width is 10 mm and stack depth is 5 mm, then you would calculate as follows:
subjectStackWidth = 10
subjectStackDepth = 5
viewAngle = 4 degrees
maximumShiftInX = tan(4 degrees) * (5/10) * 100% = 3.4963%
If you're doing the calculation in Excel, then remember to convert degrees to radians. The formula in Excel would be:
=TAN(RADIANS(4)) * (5/10) * 100
For generating stereo pairs, simply use viewAngle = (plus and minus) separationAngle / 2 .

Viewing angles that are off axis vertically use the same rules, substituting height for width and Y for X. Note that this means the same angle will require different shift percents for X and Y, in proportion to the frame dimensions.
I just stacked a dusty dried fly from my bone yard with my 10X objective. I used a typical stack depth (enough for a fly head) and my normal Zerene fly head stereo settings for a stack of that depth. I took the time to carefully measure and the stack depth by adjusting the 2 axis drilling fixture which supports my camera. This allowed me to match the travel of the actuator and using the calibrated dials on the drilling fixture I got an accurate measurement of the stack depth. I then ran the math in Rik's formulas and calculated that I was using +/- 3.6 degrees for about 7 degrees of total included angle. This is a good confirmation that my settings are definitely in the ballpark of what works for most viewers. A good sanity check of my normal settings,

And now I have a good way to "sort out" the proper settings on future stacks.

Thanks Rik!

iconoclastica
Posts: 486
Joined: Sat Jun 25, 2016 12:34 pm
Location: Wageningen, Gelderland

Re: What is a good amount of stereo separation?

Post by iconoclastica »

Interesting test. I was redirected here from a recent topic.

I can see 0-14 well; 16-24 are chaotic. 28-32 I either see as if looking through a key hole, or convex with bad registration of the details (I cannot see cross-eyed at all).
The interesting thing here is that I didn't need to zoom out to register the images. I normally have to, on this forum. When I do zoom out, the stereo becomes deeper. Without zooming, on my screen 8 is about the inside of a hemisphere, zoomed to 50% 6 degrees is. But also, even 32 degrees builds up to a stereo image (like the inside of a sausage).
--- felix filicis ---

Steve S
Posts: 83
Joined: Fri May 10, 2013 1:52 pm
Location: Southern Arizona, USA

Re: What is a good amount of stereo separation?

Post by Steve S »

8-10

rjlittlefield
Site Admin
Posts: 23543
Joined: Tue Aug 01, 2006 8:34 am
Location: Richland, Washington State, USA
Contact:

Re: What is a good amount of stereo separation?

Post by rjlittlefield »

iconoclastica wrote:
Fri Aug 27, 2021 11:19 am
Interesting test. I was redirected here from a recent topic.

I can see 0-14 well; 16-24 are chaotic. 28-32 I either see as if looking through a key hole, or convex with bad registration of the details (I cannot see cross-eyed at all).
The interesting thing here is that I didn't need to zoom out to register the images. I normally have to, on this forum. When I do zoom out, the stereo becomes deeper. Without zooming, on my screen 8 is about the inside of a hemisphere, zoomed to 50% 6 degrees is. But also, even 32 degrees builds up to a stereo image (like the inside of a sausage).
The image pairs are intended to be viewed cross-eyed. When viewed the other way (parallel or wall-eyed), there is a depth inversion that causes the ball to appear concave.

If you need to view parallel or wall-eyed, then the stereo image can be pulled into StereoPhoto Maker so the Swap Left/Right button can be pushed.

BugEZ wrote:
Mon Apr 17, 2017 5:33 pm
(formulas...)

And now I have a good way to "sort out" the proper settings on future stacks.
With recent versions of Zerene Stacker, an even simpler way is to use Tools > Calculators > Stereo Calculator, or press the "Show calculator" button at Options > Preferences > Stereo/Rocking.

--Rik

seta666
Posts: 1071
Joined: Fri Mar 19, 2010 8:50 am
Location: Castellon, Spain

Re: What is a good amount of stereo separation?

Post by seta666 »

Nice test; to me 2 and 4 do look more natural. 8 I find is too much already

Tom Jones
Posts: 325
Joined: Sat Jan 31, 2009 2:05 am
Location: Crestline, CA

Re: What is a good amount of stereo separation?

Post by Tom Jones »

6 is best for me. I see distortion starting at 8 degrees.

Tom

john w
Posts: 2
Joined: Sat Sep 11, 2021 12:56 am
Location: Gloucestershire, UK.

Re: What is a good amount of stereo separation?

Post by john w »

There is a (basically) simple answer to this. The correct stereo angle at which to record a stereo pair is the angle which will be made by the viewer’s eyes, when looking at it.

Stereo separation divided by camera-to-subject distance defines an angle, the stereo angle, such that Stereo angle = arcsine (separation/distance). ‘Arcsine’ means that angle for which (separation/distance) is the sine. You might look it up in a table, but in fact it’s easy enough to do in your head, if you can remember one example. The sine of 6 degrees equals 0.1.
So if your IPD is 65 mm and you are looking at an object 650 mm, or 2’ 2”, away, then 65/650 = 0.1 and sine-1 of 0.1 is 6 degrees. Once you have got this straight you are away. Workable stereo angles seldom if ever exceed 15 degrees, which is the eyes’ vergence at 250 mm or 10”, the least distance of distinct vision. Between 0 degrees and 15 degrees the sine is proportional to the angle, linear to an accuracy of better than 2%, quite adequate for all practical purposes. So as 2’ 2” is 6 degrees, then 4’ 4” is 3 degrees, and 1’ 1” is 12 degrees, and so on. With intermediate values between such points you could get your calculator out, or just guess it. You won’t be far out and ‘close enough’ is fine for stereo work. You will never notice the difference unless you are wildly out, which tends only to happen if you are following some fixed rule, like separation = 1/20 or 1/30 of subject distance, or so on.

So what distance will the viewer’s eyes converge at, when looking at your stereo pair? If you are lucky enough to own a true stereo lap top, probably with the nVidia system, or a 3D TV then the viewing angle can be worked out as above, from IPD divided by viewing distance. I nearly always use my laptop, viewing distance about 650 mm, so my viewing angle will be 6 degrees. For the TV 6’ might be a more likely distance so the ideal stereo angle (at 3 x 2’) will be 1/3rd of 6 degrees, that is 2 degrees. In fact many subjects recorded at 6 degrees will look OK on the TV but something that is obviously spherical will begin to look a bit ‘hyper’, like a rugby ball seen end on. For anaglyph prints (red/cyan specs., and the like) stereo viewing angle will again be IPD/distance. For an A4 print you might select 2’ as suitable so 6 degrees again; but less distance, so greater angle, for smaller prints.
With stereo viewers things are much more complicated. The situation will depend on whether you are looking exactly through the centres of the lenses (if any), or off-centre (as in Brewster type viewers) and the result with those will depend on the difference between the lens centre separation (sometimes adjustable), and your IPD. Really the only way to work this out is to do a series of test objects and see which looks best with your system, then stick to that angle. The sphere, as used by Rik, is in my view the best test object, and with the surface texture he gave it, just about perfect.

For parallel or cross-eyed viewing, I haven’t a clue. Best to do tests, as above. The brain doesn’t expect a stereo view with parallel optic axes, which are for distant vision, so stereopsis (fusing your left and right eye views into a single three dimensional perception) becomes more difficult. This may be why many find parallel viewing difficult or impossible. With divergent optic axes stereopsis is impossible, for everyone, but as this only normally occurs in coma and death it need not concern us here. Of course we have stereo perception of sorts in distant viewing, but this depends on different, monocular cues (aerial perspective, size diminution, occlusion, motion parallax etc.) and can’t be discussed here. Also it is irrelevant to macrophotography.

So why should the ‘taking angle’ for stereo pairs be the same as the viewing angle, for true, veridical, depth perception? This can be gone into in some depth but a short(ish) answer with some intuitive value is as follows. Say you are looking at a pencil, held vertically, at a short distance away. It might help here if you imagine a hexagonal pencil, with longitudinal coloured stripes. Your left eye will see further round the left side of the pencil, and the same for the right eye and right side. The exact amount of this will depend on the vergence of your eyes, at that distance, but each eye might see some (LH or RH) different coloured stripes. If you photograph the pencil, LHS and RHS, you have recorded two unique views of it. These can only be reconstructed in your brain as an object of the correct shape if you view the result with the same vergence. Say your stereo pair was recorded at 24 degrees angle, looking way round each side, and then viewed at 6 degrees. Once recorded, the LH and RH views are ‘set in stone’ but are then presented to your retinae when your eyes are converging at a different angle. Your brain knows exactly what angle your eyes are making, because it is controlling them. And the only way your brain can integrate what is presented to it is by reconstructing it as an oval pencil, seen edge on. Both the LH and RH images will contain the central, median area of the pencil, but you have given your brain too much of the sides to reconstruct as a round object of that width, at 6 degrees vergence.

It is quite possible to give a more exhaustive and satisfying account of this phenomenon, including diagrams, but I fear I am going on a bit and I hope this gives you some sort of idea. A stereo image is a construct, and is only finally reconstructed as a 3 dimensional perception in the viewer’s consciousness. You have to take the viewer into account, when recording it.

rjlittlefield
Site Admin
Posts: 23543
Joined: Tue Aug 01, 2006 8:34 am
Location: Richland, Washington State, USA
Contact:

Re: What is a good amount of stereo separation?

Post by rjlittlefield »

john w, welcome aboard! Thanks for going through the geometry aspects, and for prompting me to revisit this topic.

In many respects I am reminded of a simpler question that arises in single shot photography: "When does a lens distort perspective?"

Strictly speaking, the answer is that ordinary rectilinear lenses do not distort perspective at all.

What does distort perspective is any mismatch between the angle of view captured by the camera, and the angle of view seen by the person looking at the image. If those angles match, then the perspective as seen will be true to life, otherwise it won't.

Unfortunately, what that simple piece of theory tells us is that perspective gets distorted almost all the time.

Even with "normal" lenses like 50 mm focal length on full frame, it's not likely that the viewer of the image will be at exactly the right distance from the screen or print. With wide angle or telephoto lenses, it might be challenging for the viewer to even reach the proper position, and if they could, they wouldn't like it. After all, the point of using a telephoto lens is to make things that would otherwise span a narrow angle span a wider one. If that results in flattening the perspective by reducing scale change versus depth separation, well, that's just part of the "telephoto look". If they notice the distortion at all, viewers simply accept it as a cost of getting the main effect they care about.

I think it's kind of the same way with me and stereo. The effect I want is an immediate, strong, intuitive sense of depth -- what's in front, what's in back -- to resolve what might otherwise be a confusing mess of details. When I look out across my street with just one eye, there are several trees that merge together. I have no idea which branches and leaves belong to one tree versus another. But when I open the second eye, suddenly the scene expands to have depth, and all the structure is immediately obvious.

That immediately obvious sense of difference in depth is what I want from stereo photos.

If the stereo is weaker than it should be, I'm OK with that as long as I still immediately see the structure. Weak stereo -- the squashing of depth -- is just another version of the "telephoto look". If the stereo is stronger than it should be, I'm OK with that also, up to some vague point where I start to see stretching. The reason that point is vague is that it depends on my prior knowledge of what the subject's structure actually is. If I know that I'm looking at a sphere, then having it look like a football is disturbing. But if it might actually be a football, then exactly the same appearance is fine. Or if the structure is a gentle curve with some fine scale texturing, then maybe I'm happier to have it be stretched so as to accentuate the curve and details, even though I know that's not what it would look like if I viewed it directly.

The "BMSB highly sculptured cuticle" that I did last year says "Stereo separation is 12 degrees, which for me makes the structure more clear, at the cost of exaggerating the depth relief." Still, that's not hugely stretched, since the stereo separation for my low power scope measures a little over 9 degrees. In contrast, the stereo crystal at viewtopic.php?p=216014#p216014 notes that "The stereo separation is artificially large, resulting in about 4X exaggeration of the depth relief." Re-checking the project for the crystal, I see that the nominal angles based on magnification, step size, etc would be about plus and minus 24 degrees, so about 48 degrees of stereo separation. For a normal subject, that much separation would probably be un-fusable, even for me. But because the subject was actually quite flat, the separation turns out to be just +-4% of frame width, which is easily fusable and just results in a perception of magnified depth.

john w notes that
For parallel or cross-eyed viewing, I haven’t a clue.
Yeah, this is a complicated area.

In theory, our brains know what the convergence angle of our eyes is, and that, combined with the observed disparities, should give us a calibrated impression of depth.

But as usual, theory and practice do not coincide, and the practice is more complicated.

To take a personal example, I am at this moment sitting with my eyes located 34 inches from my monitor. My interpupillary distance is about 60 mm, varying slightly with convergence. If I were looking at left/right views overlaid on each other, the convergence angle of my eyes would be a hair under 4 degrees. However, the stereo pairs are about 7.5 inches wide on screen, so when I add the extra convergence to get crossed-eye fusion, my physical convergence increases to a bit over 10 degrees. I am old enough to be almost totally presbyopic, so focus is not a factor. Given that combination of stuff, theory would seem to predict that a camera separation of 10 degrees should look most normal. But it doesn't. Even with 10 degrees physical convergence of my eyes, it is the stereo pairs shot at 6 and 8 degrees that look most "normal", that is, the ball looks spherical. At 10 degrees the ball is clearly stretched, and more so at 12 degrees. Or going the other direction, I can pull the stereo pairs into StereoPhoto Maker, swap sides, and use a stereo viewer to fuse the images at 0 degrees of physical convergence. That does have a slight effect, in that 8 degrees definitely looks stretched. But 6 degrees still looks fine. I can even dispense with the stereo viewer and let my eyes drift outward to do the fusion. That results in about -2.5 degrees of divergence, and while 4 degrees of camera separation does look better, still 6 degrees looks only a little stretched. All told, that's a change of over 12 degrees of eye convergence, but a change of only 2-4 degrees in what camera separation looks best.

So, what's going on? I don't have a clue either. Clearly I'm experiencing some sort of "depth constancy" effect, but I have no idea exactly what's driving that. As with all things perceptual, my sensory system doesn't bother to inform my conscious mind about how it's working, only the interpretation that it comes up with. However, I'm inclined to think that other people must experience the same effect, because while I've read lots of sources that talk about having to reverse the left/right order for parallel versus crossed viewing, I do not recall reading any sources that suggested there might be strong changes in perceived depth from using the same source images in both cases. This is consistent with my usual practice of posting crossed-eye for the forum, but going to meetings with parallel prints and viewers to match. (On the other hand, the report by MacroLab3D suggests that his experience matches the theory much more closely than mine does.)

Trying to remember why I started this topic in the first place (now over 4 years ago), I think my main concerns were that (a) I wanted to be more sure that I was making stereo pairs that would work for other people also, and (b) I wanted to provide some guidance to other people about how to make images that would look good to me. I am definitely finding it interesting to revisit all the earlier replies, in light of the current discussion. I'm not sure it will change any of my practices, but it surely is interesting!

BTW, while doing some searches for related material, I chanced upon https://www.lockhaven.edu/~dsimanek/3d/ ... eality.htm . It's very interesting reading, especially if you're inclined to try making a stereo movie for viewing in a theater!

--Rik

rjlittlefield
Site Admin
Posts: 23543
Joined: Tue Aug 01, 2006 8:34 am
Location: Richland, Washington State, USA
Contact:

Re: What is a good amount of stereo separation?

Post by rjlittlefield »

Also relevant to this thread is the following experience, documented by Oliver Sacks at https://www.newyorker.com/magazine/2006 ... stereo-sue . Emphasis added...
Sue was able to see depth in the red-and-green drawings I had brought. One of these images—an impossible three-pronged tuning fork such as M. C. Escher might have drawn, with three tines of increasing heights—Sue found “spectacular”; she saw the top of the uppermost prong as three or four centimetres above the plane of the paper. Bob and Ralph, by comparison, saw it as twelve centimetres above, and I saw it as fifteen centimetres above.

I found this surprising, because we were all the same distance from the drawing, and I had imagined that a given disparity would be perceived, invariantly, as a constant depth. Puzzled by this, I wrote to several neuroscientists, including Shinsuke Shimojo, at Caltech, an expert in many aspects of visual perception. He brought out, in his reply, that when one looks at a stereogram the computational process in the brain is based not solely on the binocular cue of disparity but also on monocular cues such as size, occlusion, and motion parallax. With a stereoscopic illusion, these cues conflict, the monocular ones working against the binocular ones. The brain must therefore balance one set of cues against the other, and arrive at a weighted average. This final result will be different in different individuals, because there is huge variation, even in the normal population: some people rely predominantly on binocular cues, others on monocular cues, and still others use both. In looking at a stereo illusion such as the tuning fork, a strongly binocular person will see unusual stereo depth; a monocularly oriented person will see much less depth; and others, relying more equally on both binocular and monocular cues, will see something in between.
I would not be surprised if something like this is going on in the variation we observe between people for how much stereo separation is acceptable and/or preferred.

--Rik

Post Reply Previous topicNext topic