Hack20.Fool Yourself into Seeing 3D


Hack 20. Fool Yourself into Seeing 3D

How do you figure out the three-dimensional shape of objects, just by looking? At first glance, it's using shadows.

Looking at shadows is one of many tricks we use to figure out the shape of objects. As a trick, it's easy to foolshading alone is enough for the brain to assume what it's seeing is a real shadow. This illusion is so powerful and so deeply ingrained, in fact, that we can actually feel depth in a picture despite knowing it's just a flat image.

2.9.1. In Action

Have a look at the shaded circles in Figure 2-8, following a similar illustration in Kleffner and Ramachandran's "On the Perception of Shape from Shading."1

Figure 2-8. Shaded figures give the illusion of three-dimensionality


I put together this particular diagram myself, and there's nothing to it: just a collection of circles on a medium gray background. All the circles are gradient-filled black and white, some with white at the top and some with white at the bottom. Despite the simplicity of the image, there's already a sense of depth.

The shading seems to make the circles with white at the top bend out of the page, as though they're bumps. The circles with white at the bottom look more like depressions or even holes.

To see just how strong the sense of depth is, compare the shaded circles to the much simpler diagram in Figure 2-9, also following Kleffner and Ramachandran's paper.

Figure 2-9. Binary black-and-white "shading" doesn't provide a sense of depth


The only difference is that, instead of being shaded, the circles are divided into solid black and white halves. Yet the depth completely disappears.

2.9.2. How It Works

Shadows are identified early in visual processing in order to get a quick first impression of the shape of a scene. We can tell it's early because the mechanism it uses to resolve light source ambiguities is rather hackish.

Ambiguities occur all the time. For instance, take one of the white-at-top circles from Figure 2-8. Looking at it, you could be seeing one of two shapes depending on whether you imagine the shape was lit from the top or the bottom of the page. If light's coming from above, you can deduce it's a bump because it's black underneath where the shadows are. On the other hand, if the light's coming from the bottom of the page, only a dent produces the same shading pattern. Bump or dent: two different shapes can make the same shadow pattern lit from opposite angles.

There's no light source in the diagram, though, and the flat gray background gives no clues as to where the light might be coming from. That white-at-top circle should, by rights, be ambiguous. You should sometimes see a bump and sometimes see a dent.

What's remarkable is that people see the white-at-top circles as bumps, not dents, despite the two possibilities. Instead of leaving us in a state of confusion, the brain has made a choice: light comes from above.2

Assuming scenes are lit from above makes a lot of sense: if it's light, it's usually because the sun is overhead. So why describe this as a hackish mechanism?

Although the light source assumption seems like a good one, it's actually not very robust. Try looking at Figure 2-8 again. This time, prop the book against a wall and turn your head upside-down. The bumps turn into dents and the dents turn into bumps. Instead of assuming the light comes from high up in the sky, your brain assumes it comes from the top of your visual field.

Rather than spend time figuring out which way up your head is and then deducing where the sun is likely to be, your brain has opted for the "good enough" solution. This solution works most, not all, of the time (not if you're upside-down), but it also means the light source can be hardcoded into shape perception routines, allowing rapid processing of the scene.

It's this rapidity that allows the deduction of shape from shadows to occur so early in processing. That's important for building a three-dimensional mental scene rather than a flat image like a photograph. But the shaded circles have been falsely tagged as three-dimensional, which gives them a compelling sense of depth.

What's happened to the shaded circles is called "pop-out." Pop-out means that the circles jump out from the background at youthey're easier to notice or give attention to than similar flat objects. Kleffner and Ramachandran, in the same paper as before, illustrate this special property by timing how long it takes to spot a single bump-like circle in a whole page of dents. It turns out to not matter how many dents are on the page hiding the bump. Due to pop-out, the bump is immediately seen.

If the page of bumps and one dent is turned on its side, however, spotting the dent takes much longer. Look one more time at Figure 2-8, this time holding the book on its side. The sense of depth is much reduced and, because the light-from-above assumption favors neither type of circle, it's pretty much random which type appears indented and which appears bent out of the page. In fact, timings show that spotting the one different circle is no longer immediate. It takes longer, the more circles there are on the page.

The speed advantage for pop-out is so significant that some animals change their coloring to avoid popping out in the eyes of their predators. Standing under a bright sun, an antelope would be just like one of the shaded circles with a lit-up back and shadows underneath. But the antelope is dark on top and has a white belly. Called "countershading," this pattern opposes the shadows and turns the animal an even shade, weakening the pop-out effect and letting it fade into the background.

2.9.3. In Real Life

Given pop-out is so strong, it's not surprising we often use the shading trick to produce it in everyday life.

The 3D beveled button on the computer desktop is one such way. I've not seen any experiments about this specifically, but I'd speculate that Susan Kare's development of the beveled button in Windows 3.0 (http://www.kare.com/MakePortfolioPage.cgi?page=6) is more significant than we'd otherwise assume for making more obvious what to click.

My favorite examples of shade from shading are in Stuart Anstis' lecture on the use of this effect in the world of fashion (http://psy.ucsd.edu/~sanstis/SAStocking.htm). Anstis points out that jeans faded white along the front of the legs are effectively artificially shadowing the sides of the legs, making them look rounder and shapelier (Figure 2-10). The same is true of stockings, which are darker on the sides whichever angle you see them from.

Figure 2-10. Shaded jeans add shape to legs


Among many examples, the high point of his presentation is how the apparent shape of the face is changed with makeupor in his words, "painted-on shadows." The with and without photographs (Figure 2-11) demonstrate with well-defined cheekbones and a sculpted face just how compelling shape for shading really is.

Figure 2-11. With only half the face in makeup, the apparent shape difference is easy to see


2.9.4. End Notes

  1. Kleffner, D. A., & Ramachandran, V. S. (1992). On the perception of shape from shading. Perception and Psychophysics, 52(1), 18-36.

  2. Actually, more detailed experiments show that the brain's default light source isn't exactly at the top of the visual field, but to the top left. These experiments detailed in this paper involve more complex shadowed shapes than circles and testing to see whether they pop out or appear indented when immediately glanced. Over a series of trials, the position of the assumed light source can be deduced by watching where the brain assumes the light source to be. Unfortunately, why that position is top left rather than top anywhere else is still unknown. See Mamassian, P., Jentzsch, I., Bacon, B. A., & Schweinberger, S. R. (2003). Neural correlates of shape from shading. NeuroReport, 14(7), 971-975.



    Mind Hacks. Tips and Tools for Using Your Brain
    Mind Hacks. Tips and Tools for Using Your Brain
    ISBN: 596007795
    EAN: N/A
    Year: 2004
    Pages: 159

    flylib.com © 2008-2017.
    If you may any questions please contact us: flylib@qtcs.net