The Case For Cross

There's a battle in the stereo community - parallel vs cross. This is a strange one to me, because I think that the answer is clear - cross! There are two main reasons for preferring cross: Normal eye positioning, and ease of learning.

Before we go any further though, I should point out that those two reasons really only apply to large screens or unpracticed viewers. For small screens or people who have learned to diverge their eyes, parallel images are better.

For printed media (like the Sydney Stereo Camera Club magazine) or distribution which is likely to be viewed on a computer monitor or TV, you clearly won't be using a viewer, so surely cross would be better?

I keep my stereo pairs as wide, crossed, single-images.

But not everyone agrees that cross is better, so I'm making this page to argue the case for cross, to prompt some discussion.

---

There's a range of "normal" eye positions, based on the fact that the farthest we ever look is the stars, and the closest is the tips of our noses - that's a range between parallel eye-lines and crossed-eyes:

There's a very simple consequence to this: The centres of your eyes are about 6.5cm apart, so for the parallel case, the images you're looking at can't be more than 6.5cm apart either.

If you want to look at a image with a baseline higher than 6.5cm - say a photo on your computer monitor - then cross is possible, easy even:

But for parallel, you have to force yourself to go "wall-eyed", which isn't in the natural range of eye motion, and most people can't do (though it is possible to learn eventually):

You may pick on my wording: "Most people can't do". Really? Yes, I think really. You can find some stereo enthusiasts holding images at arms length, warping a page, or just giving up when a baseline is too wide. Going wall-eyed doesn't correspond to any natural thing that you might do, and I'm quite sure that I could make a case for it causing more discomfort.

For cross on the other hand, it's easy to do wide baseline. Although it's more comfortable to move back a little, I can easily cross-view images at full screen on my 48cm-wide monitor, sitting about 60cm away. In fact, I often do. That's a 24cm baseline at close range! I challenge the parallel-preferrers to try just half that separation - 12cm - at any range which still allows them to see the image clearly.

Learning to cross has yet another advantage: It's easy to teach! Just use the focus-on-your-finger method (see page):

So the case for cross is that simple!

I'll cover a couple of counter-arguments in advance:

"Magic-Eye" pictures use parallel, not cross, because with parallel you see the depth of the image to be inside the window which is natural, instead of floating in front of it. The same argument applies to stereo images.

I don't believe so. Magic Eye pictures have a very small baseline, and you keep the image context in view, so the framing matters. When you "free-view" an image (parallel or cross with no viewer), you lose the context of the rest of the world around you, it's often hard to even tell if the image is inside or in front of the stereo window, let alone the position of the window compared to the surroundings.

There's less eye-strain viewing in parallel than in cross.

Okay, that's not a real-sounding question, that's just a Dorothy-Dixer to let me address this issue.
Eye-strain would be a measure of how different the focus is from the vergence - with real 3D objects and well-made 3D movies, these agree, and there's no eye strain. For free-viewing images, the focus and the vergance are different, which causes eye-strain (or, some claim, is the main cause but not the only cause).
Answering the question of eye-strain though, depends on deciding that if something's actually 1m away (hence you focus 1m away to look at it), does it cause more strain to verge on infinity than it does to verge 50cm away? I'm not sure how to judge this, because even going through the math for a logarithmic focus range, different stereo separations, and different distances from the eyes, you'd need to run a study on which is better/worse in terms of human comfort if all else is equal. I'd love to hear if anyone has some good data or meaningful rules of thumb. I suspect that the problem is just too poorly defined, so the any answer will be too long or will carry too many caveats and conditions to be generally useful.

If parallel is easier when you're using a viewer, shouldn't we just standardise on that?

Not if it's worse!
I think, surely you know if you're going to be using a viewer, because you've printed your image on a viewer-card. If your image is online or printed in a magazine which is awkward to put in a viewer, why not go for the more appropriate format?

But I already know how to parallel-view, learning to cross-view is extra work!

That was my situation about 3 years go, when a friend pointed out that cross would be easier when he was messing with some spot-the-difference images for a puzzle. I used to get frustrated with baselines over a couple of cm, but I found instructions online, and learned how to cross-view in just a couple of minutes - it was really easy, and I had it mastered in a few days. Learning for the first time is tricky, as with anything, but if you've already learned how to "snap" that image into place and "focus wrong" for parallel, you'll find that learning to cross-view is really simple.
I made a guide to help you learn - try it!
Now that I can do both, I find cross easier, and much faster to snap to.

Okay, this already contains way too much text, I'll stop here, and address further issues below if anyone raises any.

---

The Case Against Cross

Okay, okay, so I went in a bit strong. It turns out that there is a very, very good argument for parallel viewing sometimes, and that's the apparrent magnification. I'll leave it in the original words of a response:

re your web page "The Case For Cross"

Yes, I can parallel-view stereo pairs with 12 cm separation from 60 cm. Admittedly, it's a bit of a strain. But I find parallel-viewing of stereo cards with normal separation effortless. Cross is less natural for me.

For orthoscopic viewing (did I just invent that word?), the eye should be on the optical axis of the photograph which, assuming it's not cropped, would be the line perpendicular to its centre. For most image sizes, the eye would diverge much further from this line for cross, compared to parallel viewing. When the image width equals the eye separation, the eye is actually on the axis for parallel viewing, which it never is for cross viewing. This must cause some warping of the 3D space.

The clincher for me, however, is the apparent size of the stereo image. For cross, this must lie entirely in front of the image, for parallel behind the image (excluding anomalous extreme images where corresponding points cross over to the 'wrong' side of the original pair). So with cross viewing you feel you are looking at a tiny miniature model of the world. For parallel, you feel you are looking through a window at a full-size real world.

Regards,

Chris

I'll leave the 3D warping for now because it's an interesting point that I'm unsure about, but Chris raises a point that I noticed when using Disparity:

At the same (relatively close) viewing distance, an image feels almost twice as big in parallel as it does in cross

This is more than just a slight illusion: On one test picture, I couldn't really read some text in cross because it was too small, but I could in parallel! It's a matter of your brain's expectation of distance vs. the actual image distance, I don't think that there's any appreciable optical magnification (as you see with changing focus in a camera).

On that point alone, I will now view small images in parallel when I have the choice, but I still prefer large images in cross.

-- Peter.