Not very well, apparently

February 10, 2005 3:57 PM

How do we see?This site by Dr. Dale Purves makes it obvious we don't see things like a camera in any way. Check out the interactive demos, test your perceptual abilities, and read the research explaining why this happens. Number 12: Color Contrast Cube is particularly startling. Warning: Totally Flash interface, but appropriate for subject matter. More experiments at a less Flash-y associate's site.
posted by JZig (19 comments total)

Sigh, second link should be this
posted by JZig at 3:59 PM on February 10, 2005

Neat stuff. I found the 'color constancy: cube' rather interesting.

The fact that we see perceive the tiles as 'blue' or 'yellow' when they are technically 'gray' is a feature of the human vision system, not a bug. We intuitively pick up on cues from the image to determine the color cast, then work backwards to determine what the 'true' colors of the squares are.

In computer vision applications, it is sometimes desirable to use an object's color to help identify it. Although the object's color seems constant to us, to a machine it can vary quite a bit depending on how it's lit (sun, clouds, indoor lighting, etc.) We automatically correct for this and determine the object's real color, but getting a computer to do this is quite difficult and is still an open research problem.
posted by driveler at 4:20 PM on February 10, 2005

If you consciously know about these limitations you can do pretty well. On the contrast tests I can either try to block out the surroundings mentally or at least make a (usually pretty accurate) guess as to how they match. Computers start with the raw data and have to figure out, say, the amount of illumination of a given side of a cube in order to subtract that and get the real color of that side. Our conscious minds have the opposite, we get a very well-processed image that's already done that and we don't really have access to the original raw data, but we can extrapolate it, often better than computers currently can go the other way.
posted by abcde at 4:36 PM on February 10, 2005

I love this kind of stuff.
posted by Toecutter at 7:26 PM on February 10, 2005

Purves wrote the book on neuroscience, literally. I didn't know he had moved on to psychophysics.
posted by euphorb at 7:28 PM on February 10, 2005

"Although the object's color seems constant to us, to a machine it can vary quite a bit..."

Color in machine vision is a problem because color is subjective and not objective. To correctly identify color (where human color vision is normative), you'd need to duplicate the functionality of the whole biological system—from the cones to the brain—that creates the perception of "color".

While there's a lot of subjectivity in every part of human vision (as said above and in the link), an object's shape is far more objective than is color.
posted by Ethereal Bligh at 8:13 PM on February 10, 2005

Yeah, the contrast ones didn't work so well for me, especialy the monocrome ones. I've seen these before and I guess my brain has learned to 'unsceneify' the images.
posted by delmoi at 9:20 PM on February 10, 2005

I know if you take a picture in a supermarket, the picture has a green tint. This is because that's how the supermarket with its flourescent lights really looks. We filter out the green, because we know it shouldn't be there. I remember this was kind of a surprise for me in my photography class.
posted by xammerboy at 9:48 PM on February 10, 2005

Thanks; I get a kick out of this stuff (too).
posted by ikkyu2 at 10:17 PM on February 10, 2005

Great stuff!
posted by ParisParamus at 10:25 PM on February 10, 2005

This is good!
posted by orthogonality at 10:31 PM on February 10, 2005

For some reason, I'm absolutely horrible at attempting to change my perceptual system consciously. So, I know rationally based on how the examples are setup exactly which colors will end up the same, but I can't actually experience it. I'm still not sure if this is important or not, since I can base my decisions on my rational knowledge. Anyway, that's what the title meant :) Picked up this link from my class on Sensation and Perception. Everyone should take a class on perception, the material is fascinating.
posted by JZig at 11:46 PM on February 10, 2005

The "chromatic adaptation" demo kicks...ass.
posted by symphonik at 12:18 AM on February 11, 2005

I know if you take a picture in a supermarket, the picture has a green tint. This is because that's how the supermarket with its flourescent lights really looks.

Another example, familiar to any skier who's ever worn strongly tinted orange goggles: when you first put on the goggles, everything seems very orange--but for a very short time; I find my mind adapts and everything seems the "normal" color again within 30 seconds. Even more striking, when you take the goggles off, everything seems very blue (as your mind is still compensating for the orange goggles which are no longer there) for an equally brief period.
posted by DevilsAdvocate at 7:22 AM on February 11, 2005

Cool stuff! My father, author and illustrator of another Sinauer book on vision, would have loved this.
posted by rodo at 8:24 AM on February 11, 2005

Wow, your dad illustrated the Rodieck book? That's some good work. Practically every page has some nice color figures.
posted by euphorb at 9:19 AM on February 11, 2005

Fascinating stuff as always. What, nobody's posted the green-cylinder-on-the-checkerboard animation yet? I would, but I can't remember which threads it's in, and I guess everybody's seen it by now. Plus I'm incredibly lazy. Anyway, I loved the one that's like it on here, with the slider control so you can watch it change from "brown" to orange.
posted by soyjoy at 9:40 AM on February 11, 2005

I would if I could.
posted by breezeway at 1:14 PM on February 11, 2005

I immediately thought of the checker board illusion too...It blows me away every time I look at it.
posted by tdstone at 1:03 AM on February 12, 2005