How far can stereo take you?

[Zloth]

Just how far can people judge distances by eye triangulation alone? I'm sure it's different from person to person but I'm curious about roughly how far it can work without the help of all those other things our brains do to figure out distances.

[cybereality]

With a 24" monitor I can see about 1 foot in front of the screen (in a best case scenario). Depth may be further in, but still around 1 foot inside the screen. Of course, with the larger screen you can get more depth (for example on a 50" HDTV its much better).

[WiredEarp]

I think hes meaning how far do objects have to be away before your eyes go parallel and no longer provide any depth cues due to their convergence.

I have no idea, but 6-10m would be my guess.

[Zloth]

Yeah, that's what I'm looking for.

Say you're floating out in space and there are a bunch of featureless, white balls floating around you. Some are close and some are far but each ball's size is such that they all appear to be the same size from where you're floating.

You're going to have no problem telling a small ball 3m away from a little larger ball 6m away. Obviously you aren't going to be able to tell a ball that's 3m away from a ball that's 3.0001m away but I'm looking more for the point at which your eyes are so close to parallel that you won't be able to tell which ball is further away even when one is twice as far away as the other. Can you tell a ball that's 10m away from one that's 20m away? 20m and 40m?

[WiredEarp]

This looks like what you are after (from wikipedia):

Convergence - This is a binocular oculomotor cue for distance/depth perception. By virtue of stereopsis the two eye balls focus on the same object. In doing so they converge. The convergence will stretch the extraocular muscles. As happens with the monocular accommodation cue, kinesthetic sensations from these extraocular muscles also help in depth/distance perception. The angle of convergence is smaller when the eye is fixating on far away objects. Convergence is effective for distances less than 10 meters.

So, I guess my 10m guess wasn't too bad

[Okta]

The only way to know would be trials using a distant object against a void so there is nothing to judge by. 10m doesn't sound like much but i wouldn't doubt it.

[Fredz]

This study (Depth perception in computer graphics) seem to confirm that stereopsis is the most important depth cue for up to 10 meters if there is no motion parallax (ie. head or scene not moving, less than 10 m else). For higher distances other cues are more effective for depth estimation (texture, then brightness, size, etc.).

Depth perception.png

But your first question seemed to be more concerned by absolute distance estimation rather than depth estimation (ie. relative distance), and binocular disparity alone can't provide absolute distance perception. For that you need accommodation and convergence as explained in the Wikipedia entry about depth perception :

Of these various cues, only convergence, accommodation and familiar size provide absolute distance information. All other cues are relative (i.e., they can only be used to tell which objects are closer relative to others). Stereopsis is merely relative because a greater or lesser disparity for nearby objects could either mean that those objects differ more or less substantially in relative depth or that the foveated object is nearer or further away (the further away a scene is, the smaller is the retinal disparity indicating the same depth difference).

In stereoscopy accommodation and convergence are not correctly emulated since the eyes always accommodate at the same distance (the screen plane), so I don't think it's possible to perceive absolute distance in this case.

[cybereality]

Ok, I see what you are asking now. In my guesstimate I would say past 100 feet (30 meters) there's not any 3D.

[Zloth]

THAT was a very interesting dissertation! Thanks Fredz!

So 10m with stereoscopic being a dominant factor but it looks like it's something of a factor out to at least 50m. That's a mighty long triangle. I doubt our screen resolution is going to allow anything close to that. So our monitors are probably going to be limitting our depth perception more than our natural abilities.

There's a lot of other interesting info in there. I've been pushing my FOV down in games when I can just so the moon doesn't look like a big egg in the sky but, now that I think about it, maybe I should do that even more than I have been for S3D.

[cadcoke5]

I suspect that user head motion is able to greatly increase the distance you can perceive distance. I call this "head bobbing", or "Prairie dog effect." It may be related to stereoscopic vision, since you are essentially creating a wider set of eyes, though it is not true convergence. Perhaps the brain is able to use the same "subroutines" as it does when processing convergence information from the eyes.

With the use of camera-based head tracking being implemented with gaming devices I suspect the lack of special glasses or head-ware will make it popular.

Joe Dunfee

[Fredz]

This "head bobbing" you are refering to is actually what is called "motion parallax" in the previous picture. It's less effective than stereopsis for depth estimation for distances under 10m but a little bit more effective for longer distances up to 100 or 200m it seems.