a try at that Spanish Castle thing

[eub]

(Shocking-to-me perceptual trick ganked from here.)

I gave it a try and mine sort of works; the indefinite persistence is there, but it doesn't manage full color saturation. It might just be my choice of source image. I could try oversaturating my inverse-color image, but it would clamp very soon. I simply set the inverse-color image's luminance to 50% everywhere; I didn't check to see if that's what the "Spanish castle" one did.

What I'm really fascinated by is what characteristic of the grayscale information of the image is needed to make the color afterimage persist. It can't possibly work with a flat gray field, right? But it does work with the castle's sky. So maybe it works with a grayscale gradient, or with a flat area alongside something else in the frame. I should try some of those.

Comments

[rehana.livejournal.com]

Having looked at some of the other examples that page links to, I think the key is having an image that looks right even when it's not very saturated.

[beaq.livejournal.com]

The castle image doesn't look any more saturated than yours. And I saw a picture from Arches N.M. that only shows the blue and not the very red color that I know the arch has. Also, I think you can't look directly at the blue area and have it work -- maybe a blind spot issue, maybe something to do with color receptor arrangement?

I think yours works as well as any others I've seen -- it's just the castle has a lot of high-contrast areas, where a little color goes a long way.

[meerkat299.livejournal.com]

A few pointers to make it better, mainly when choosing a source image:
--as beaq mentioned, a lot of high-contrast areas, where a little color goes a long ways
--A great way to accompish the above is to choose an image in which the center of focus is surrounded by a low-detail (low value contrast) field of a complementary hue. The castle image uses this technique with the amber hue of the stonework contrasting with the blue of the sky. This makes both colors stand out better.
--The greyscale image should have a lot of high-value (whiter than medium grey) areas of color; the since you cannot move your eyes from the dot, it is only the lighter and brighter (high-value, high-saturation) areas that will be visible in the peripheral areas.
--Important: the dot should be on the center of focus (high-contrast/detail) for the picture, otherwise your eye will be drawn away from the dot, destroying the effect. Strong colors, especially reds, will attract focus and should only be used carefully: if reds are used, keep them near the dot.
--keeping the previous point in mind, the dot should still be close to or at the center (as you have done); an consistant amount of periphery is less likely to draw the focus away from the dot.

complementary color pairs in RGB: blue/amber; purple/green; red/aqua

[katybeth]

I wonder how much it helps to choose an image for which your brain know what the colors are "supposed" to be (e.g. sky is blue).

[bhudson.livejournal.com]

When I keep staring directly at the dot, I clearly notice the colour fading out. It takes about 5-10s before I lose the illusion entirely. Different colours disappeared at different rates: the castle was greyscale very quickly, the grass was second; and it took several more seconds of the sky slowly getting grey pixels and then large grey patches (radiating out from my focal point) before it was all grey.

Neat!

[aquaeri.livejournal.com]

I have no clue what's supposed to happen when I move the mouse. Nothing changes. I get a pretty strong inverted (correct colour) afterimage if I look at something white, but not if I keep looking (that's what I'm supposed to do, right?)

[bhudson.livejournal.com]

The mouseover switches to a greyscale image of the negative you were just looking at. The illusion I see is that you get a positive-colour image.

[aquaeri.livejournal.com]

Ah, no, mouseover is doing nothing. This is possibly because the net currently is slower than snails - it's taking quite a long time to download the inverse-coloured image.

If you get a grey-scale replacement for the inverse, I'm not remotely surprised at the illusion - you're seeing the afterimage on top of something that's giving you all the fine details (which afterimages tend not to be good at.)

[bhudson.livejournal.com]

Ah; well, preload the two images; then run the experiment and it should work. Even if it doesn't surprise you, it's kind of neat.

[eub.livejournal.com]

It's javascript-based (and the js is supposed to preload the images); do you have that turned off?

What's surprising is that the chroma afterimage when fused into the grayscale seems to persist dramatically longer than an afterimage ordinarily does. Then if you shift your point of fixation it breaks and is gone.

[aquaeri.livejournal.com]

I figured some javascript out, but now I've killed the page and I don't want to go back to reloading it (we ate our quota for the month, so no images on the net until Tuesday). So, I didn't get the castle working for some reason, but I did manage to get yours - except I only really got the pink of the flowers on a near-grey background. I think someone else mentioned having a colour contrast pair near the focus spot and I would guess that might help.

I didn't get the afterimage to last surprisingly long. But since a lot of people mention it, I wonder if this combines a normal afterimage colour effect with another one. I can't remember what it's called but I read about it somewhere like Scientific American, I'd guess late 1980s. You'd look at horizontal black-and-magenta stripes and vertical black-and-green stripes, and then when you looked at horizontal black and white stripes, they would fill in with green (and vertical with magenta). It wasn't just afterimage, it'd actually last for up to ten minutes - you could go away and do something else and come back to the magazine and yep, the horizontal stripes still looked greenish. Given the greyscale image used here, it sounds related.

[eub.livejournal.com]

McCullough effect -- pretty amazing thing. When I first tried that (I think the same Sci Am article) it lasted faintly for days; I was afraid I had bent my brain.

This one is related at the very least in that it's got some of the same kind of color/texture "contingency" going on. The thing that first strikes me as different about this one is that the effect disappears and doesn't come back as soon as you break the luma/chroma alignment, whereas the McC effect you can put through the wash with no harm.

[aquaeri.livejournal.com]

It makes sense to me that the McCullough effect (thanks for remembering for me!) would last longer - the stimuli (even width black stripes) is much simpler. You'd be asking whatever part of the brain 'remembers' the green stripes to remember something far more complicated, in terms of colour and placements and shapes. Also, during the "training" part, you're not seeing the black and white part of the pattern, unlike McC.

At least, that all sounds like plausible make-it-up-as-I-go reasoning to me :-).

[blk]

Thought you may be interested: another of my friends gave this a try.

[eub.livejournal.com]

That's just not a friend, that's a jcreed!
Thanks.