Ok, someone with a little training in eye anatomy help me out here...

[Doodler]

http://www.johnsadowski.com/big_spanish_castle.html#

[farmerjumperdon]

I think it is the brain that's being fooled.

My experience with brain anatomy is somewhat limited, confined mostly to examining road kill. My soul conclusion is that it is mostly made of bluish gray gelatinous stuff, . . .and that I'd like a new one.

[Vermonter]

Wow!

[Nicolas]

Did you know that only the central focal part of your vision is in colour, and the peripheral part is basically b&w? the brain fills in the colour info needed to give a seemingly fully coloured vision. and that's why it remains in colour until you move your eyes. The brain assumes the colours in the peripherical part to be constant as long as it has no new info. So until you move your eyes, you keep on seeing the colours there were before, giving the colour image when added to the B&W. Apparently there is some shift (even inversing) because it seems to be no simple addition of colour and B&W.

Perhaps the long exposure alters the sensitivity of your receptors, making them less sensitive for those colours and hence inversing the colour scheme when these colours are suddenly gone.

But anyway, the basis is in that the brain fills in the peripheral colour info of your vision because your eyes can't see colour in the periphery.

[Argos]

Ok, you´ve got me. I would explain that, but I´d surely spoil it all. I know the explanation the fellows are giving are facetious, of course.

[Ronald Brak]

Cool!

Let's see, how can I quickly explain what is going on? The cells that detect colour work in opposition to each other. If you tire one type of cell out by staring at a colour without moving your eyes those cells will get tired. But the cells that are working in opposition don't get tired so that when you look away the cells that are in opposition are working harder than the tired cells. This results in you seeing colour when their isn't any. The tired cells are no longer cancelling out the colours the other cells detect.

Someone else could probably explain this better.

[Nicolas]

The effect you explained, is the explanation why the colour info is inverse Ronald.

My info on periphery explains why colour remains until you move your eyes.

So it goes like this:

*stare at the dot. You only see colour around the dot, the rest is filled in by the brain, largely remembered from when it was in focus for a brief moment (during the picture scan you did).
*the long exposure tires the receptors for those colours, making you less sensitive to them.
*when these colours go away, you see the inverse colours.
*your brain gets this colour info and uses it to fill in your peripheral view
*only when you look away, the brain sees the real colour info that is now in the new focus point (anywhere on the pic other than the dot, or even outside the picture), and sees that the image in fact is B&W.

There are some small gaps in this explanation, as it would require the brain addition of colour to get tired (or decrease its gain) as well. Or maybe the very few colour receptors in your periphery (or the very brief scanning motions of your eye) make the receptors there tired as well, giving rise to the inversing action in the periphery as well.

But the fact that the periphery is B&W needs to be a factor, it can't be only tiring (overexposing) your receptors as the time to B&W depends on when you move your eyes. Move them immediately, and it goes B&W within one second. Wait longer, and it will remain there very long.

[Nicolas]

So what I think he does to make it:

*take a colour pic (and add a dot :)).
*separate B&W and colour info.
*inverse colour info.

Now you need to stare at the inverse colours first. Then the B&W info is added when you move the mouse over it.

[Argos]

Explanation [highlight the phrase bellow with the mouse]

It´s just a javascript rollover

[Nicolas]

It´s just a javascript rollover

Argos: no it's not. A Javascript rollover cannot see when you move your eyes. Obviously the switch of the primary strange colour thing to the B&W image is javascript, but there never is a real colour pic shown on screen.

[Argos]

Nicolas, such an illusion cannot exist.

<a href="#" onMouseOut="MM_swapImgRestore()" onMouseOver="MM_swapImage('big','','2006_stuffs/manzana2.jpg',1)"><img src="2006_stuffs/manzana1.jpg" name="big" width="720" height="495" border="0"></a></p>

In bold, the two images the author of the trick uses. The event on MouseOver calls the function MM_swapImages.

[Nicolas]

Yes OF COURSE there are two images. The strange colours you see first, and the B&W in the end. Your explanation requires 3 images, namely also a natural colour photograph.

Check it out yourself, there never is a natural colour jpg into view, and the illusion disappears only at eye movement so that can't be triggered by code.

Or read up on peripheral vs foveal view and colour receptors, that might be informative too.

[Argos]

Don´t be self-deceptive.

Edit: And don´t misguide other people.

[HenrikOlsen]

:)
Optical illusions are by definition selfdeceptive, the interesting thing is what they tell about the mechanisms of vision, in this case about coloradaption.

If you load and try to repeatedly reload the second image(on top of itself) you'll see that there is absolutely no color information in it, it's not an animated .gif that fades or anything silly like that.

The interesting thing to to figure out why it's seen as a full color image for a short time after loading when you've been looking at the first image for a while first.

[Lance]

Nicolas, such an illusion cannot exist.

<a href="#" onMouseOut="MM_swapImgRestore()" onMouseOver="MM_swapImage('big','','2006_stuffs/manzana2.jpg',1)"><img src="2006_stuffs/manzana1.jpg" name="big" width="720" height="495" border="0"></a></p>

In bold, the two images the author of the trick uses. The event on MouseOver calls the function MM_swapImages.
Oh, this is funny...

No, there is no trick. It is a real illusion.

Here, look at the two pictures seperately:

http://www.johnsadowski.com/2006_stuffs/manzana1.jpg

and

http://www.johnsadowski.com/2006_stuffs/manzana2.jpg

There is never a color image displayed.

[HenrikOlsen]

Well, there is, the first image is in color. :)

The colors are just inverted the right way so that local whitepoint adaption will overcompensate and show colors when it's gone.

Incidentally, just to avoid confusion, you need to have javascript enabled to see the illusion as it depends on switching to a second image.
It took me a while to realise that after waving my mouse around over the first image.
Incidentally that gave a strong clue to how the illusion works, since after looking on the dot for a while the colors looked faded but came back strong in the vicinity of the mouse when I moved it in front of the image.

[Nicolas]

AAAAAAAAAAND your pheriphery cannot see colour, it's filled in by the brain. And that's why the colour illusion remains until you move your eyes, not merely until your eyes have recovered from the long exposure.

:)

[Nicolas]

Don´t be self-deceptive.

Edit: And don´t misguide other people.

And most important of all: don't be overskeptical, or overconfident... ;)

[Argos]

OK. See a little flag over the tower on the left? I´ve taken a sample of that region on the two images. On the "B&W" one the RGB values are 70,70,70. On the "color" one the values are 89, 134, 93. So, the images are equal?

[Lance]

The illusion has you so completely fooled that you don't even realize it's an illusion. The author should be proud.

[Argos]

Ok. NOW I understand it all. Excuse me, I didn´t catch.

Looking with care I can see what the ilusion is about. There is a full color image appearing briefly. I didn´t see it at first.

I thought that you guys were insisting that the first image was Black and white. I´m sorry.

I owe you an appology, Nicolas.

[Nicolas]

No problem. :)

Try to focus on the dot for 30 seconds, then move the mouse into the image but KEEP looking onto the dot as long as possible, don't move your eyes at all. The colour image will remain very long. I had it there for more than 10 seconds before I got tired of it :). You very, very slowly see it becoming just a tiny bit less coloured that way.


I had problems with overexposure in the simulator: I couldn't see the red line moving on the blue screen after it had been still for a long time. I'll propose a B&W screen some time...

[Nicolas]

As I said, that "short while" can be really long if you just hold your eyes focused on the point all the time, also after you moved the mouse. I think that 30 seconds and more must be possible, but I haven't got the patience to check it :).

[Argos]

Yes. :) The color image persists for almost 1 minute if you keep locked on the point. It begins degrading from the center and the blue sky gradually loses color.

The brain somehow stores the color information from the first image and combines it with the second B&W image in a sorta CMYK mode.

[ToSeek]

I managed to watch it long enough to have it fade from color to black-and-white. Boy, that was strange!

[Peptron]

Even worse: If you look at the dot while you have the color illusion going on, if you move away you'll see black and white. BUT, if you quickly come back at the dot, the colors will be back!

[TheBlackCat]

The periphery does have color vision. It may not be as acute as in the center of your vision (the fovea), but it is still there. The center of your vision does not have the low-light rods, your periphery does. But your periphery does have some color-sensetive cones. There are no where near as many color receptors, but there are most definitely still some there.

This really has nothing to do with your photoreceptors getting "tired". Cones can't get "tired" like other neurons, they have a specific adaptation called a "ribbon synapse" that prevents this. They can continue releasing neurotransmitters indefinitely. In reality, looking at an image cannot make your receptors tired, receptors are active in darkness and shut down when light is present, if anything they would be resting when looking at a bright computer screen.

What we are really dealing with here is a phenomenon called "simultaneous color contrast". What happens is that your eyes are set up to filter out changes in scene frequency content. For instance, an apple looks like an apple when out in the sun, in the shade, under fluorescent light, and under incandescent light. In reality the frequency content of the light hitting your eyes is completely different. There are two ways this is accomplished. First, the visual system is set up to filter out changes in the overall frequency content of the scene. So the scene looks largely the same no matter what the frequency of light falling on it is (within certain limits). This happens by supressing cones that are sensetive to a color that is overabundant and reducing supression of cones that are sensetive to a color that is not abundant.

That is not the whole story of what is happening here, although it likely plays a large role (more on that later). What is happening here is what is called "simultaneous color contrast". As was mentioned before, different ganglion cells (the neurons that connect your retina to your brain), are sensetive to different pairs of colors. It may be best stimulated by a red dot surrounded by a green donut or vice versus, or a blue dot surrounded by a yellow donut (or annulus) or vice versus (the dots and donuts being made up of just a handful of cells each). The opposing color to a region's target color falling on that region will inhibit it. For instance if the cell is sensetive to a blue center and a yellow surround, either blue hitting the surround or yellow hitting the center will inhibit it. Thus blue hitting both will cause them to cancel out, as will yellow hitting both. A scene that is uniform in one color will not stimulate these cells at all because the stimulation by regions sensetive to a color will be cancelled out by inhibition by the surrounding regions inhibited by that color. (note these regions are pretty small, a handful of cells). This is called "color opponency".

However, when you combine this with the fact that the visual system tries to correct for changes in scene illumination, it is thought that you get "simultaneous color contrast". Although simultaneous color contrast is known, the exact mechanism is not clear. However, here is probably the most likely reason. The output of the receptors in your visual system adapt to changes in frequency content in the environment in order to maintain a constant color to the environment (note that the receptor electrical activity remains the same, only their output changes). In the image you saw, there is an overabundance of some colors in some areas. Your visual system corrects for this by reducing the output of the receptors sensetive to those colors. Under normal conditions this would correct for gradual changes in spectral content in a scene. This supression does not change instanteously, it takes time (this makes normal, gradual color changes more subtle and thus less noticable).

What happens when you suddenly eliminate those frequencies from the visual scene is that the receptors sensetive to those colors are still being supressed for a period of time. However, suddenly you have a scene with all or no frequencies present (black or white). That means that the colors you were looking at are being supressed, but the colors you were not looking at are being detected normally. Due to the opponency effect, the receptors that were sensetive to the colors in the original scene, and thus are opponent to and thus supressing the colors not present in the original image, are being supressed themselves. Since they are no longer active, they are unable to supress their opponent colors anymore so you see an overabundance of those opponent colors. The colors that were not present, however, now are present, allowing them to supress the already partially-supressed original colors, making those parts of the spectrum disappear. What you are left with is the colors that were previously present being supressed, while the colors that were previously absent being unsupressed (thus enhanced).

Similar things happen with black and white, due to overall contrast as opposed to individual color contrast.

Now at least in Cyprinid fish, the only one I am aware of with a working computational model on the subject, it appears that this is mostly operating in the earliest stages of the retina, with a network called the horizontal cell network directly altering photo receptor output both for color contrast and color opponency. It is not clear whether this is the case with primates or not, but considering how similar all vertebrate retinas are it very well might be. The phenomenon itself is certainly the same, even if the cause is not.

Now I should point out one thing. In reality when I say a receptor is supressed or not being stimulated, what actually means is it becomes more active. A receptor that is being stimulated becomes less active. This may seem counterintuitive, but photoreceptors are shut down by light, or by the color they are sensetive to, and become more active in the dark. Your retina is actually the most active when you are asleep. I didn't use the correct terminology through this post because it is not important in this case and makes things hard to follow, but it is something to keep in mind.

[TrAI]

Hmmm... My guess is that when looking at the image with the (complimentary) croma information, the eye will lose sensitivity to those colors in those areas due to depletion of photosensitive chemicals in the coresponding cones, thereby gaining sensitivity to the complimentary colors of this. Then displaying the luminance image will mean that the eye will se the hues it is now more sensitive to. As long as you do not move the eyes, the grayscale of the second image will deplete the cones of all colors at a similar rate, and so the relative depletion will remain for a while(of course, as a cone get more and more depleted, the rate of depletion will decrease, and the previously less depleted cones will catch up, so the effect fades) . Moving the eyes destroy the effect because the persistant image caused by the depletion and the real image no longer overlap, my guess is that the brain is wired to suppress persistant images, there is probably no natural circumstances where you can get the effect caused by this illusion, and so the brain never learn to suppress persistant color caused by depletion when the color patern matches what you currently see. Since you have moved your eye, the effect that would keep the illusion persistant longer(the balanced depletion I mentioned earlier) will also be broken.

I do not think the effect that the brain fills in color is a major cause of the illusion. The eye can detect color in the entrire field of view, it isn't colorblind everywhere but the center of the field, however, it is true that the fovea have the highest density of cones, and the color sensitivity drops of the further from this you get. If you do not look at the spot in the middle of the image, but outside the image, you would still get the same effect when switching images, it is important to keep the eye still, it will perhaps seem a bit unfamiliar to keep your attention on something outside the sharpest part of your field of view, and your eyes might want to move towards the object of attention.

[TheBlackCat]

Hmmm... My guess is that when looking at the image with the (complimentary) croma information, the eye will lose sensitivity to those colors in those areas due to depletion of photosensitive chemicals in the coresponding cones, thereby gaining sensitivity to the complimentary colors of this. Then displaying the luminance image will mean that the eye will se the hues it is now more sensitive to.
Your retina is quite capable of keeping the pigment level constant, at least for cones. That is not the issue here. It has to do with your retina trying to make sure you percieve the same colors regardless of the frequency content in the environment. If the pigment level gets depleted you go blind (that happens with rods in anything other than very dark environments). That does not happen with cones, in fact your cones will be destroyed long before you get enough light intensity to seriously overwhelm the refilling mechanisms (assuming nothing disrupts retinal metabolism, the low pressure and thus low oxygen in an airliner is enough to dim vision somewhat).

Moving the eyes destroy the effect because the persistant image caused by the depletion and the real image no longer overlap, my guess is that the brain is wired to suppress persistant images,
This is not happening in your brain, at least not entirely, much or all of it is happening in your retina (that much is clear, at least it is for lower vertebrates whose eyes are pretty much the same as ours). And your retina does not supress persistant images, it supresses changes in color content or overall ambient illumination level. Moving your eyes too much (it is impossible to keep your eyes perfectly still) means the receptors are getting changing frequency contents and thus don't have a chance to adapt to one specific set of frequencies.

[HenrikOlsen]

So it is as I said that it's the whitepoint compensation that's fooled.

[TrAI]

Your retina is quite capable of keeping the pigment level constant, at least for cones. That is not the issue here. It has to do with your retina trying to make sure you percieve the same colors regardless of the frequency content in the environment. If the pigment level gets depleted you go blind (that happens with rods in anything other than very dark environments). That does not happen with cones, in fact your cones will be destroyed long before you get enough light intensity to seriously overwhelm the refilling mechanisms (assuming nothing disrupts retinal metabolism, the low pressure and thus low oxygen in an airliner is enough to dim vision somewhat).

This is not happening in your brain, at least not entirely, much or all of it is happening in your retina (that much is clear). And your retina does not supress persistant images, it supresses changes in color content or overall ambient illumination level. Moving your eyes too much (it is impossible to keep your eyes perfectly still) means the receptors are getting changing frequency contents and thus don't have a chance to adapt to one specific set of frequencies.

Hmmm, I was not actually trying to say that the cone can not regenerate quickly enugh, the reduction is related to the cell trying to adapt to the level it is exposed to, you need less of the stuff when looking at something bright than something dark, I would think that regenerating so much of the proteins that you saturate the cells response would make you just as blind as loosing it all.

I realize now that using the word "depletion" may have been a poor choice of wording, I see it is usualy used about rapid and complete emptying of something, perhaps reduction or lowering or some other word would have been better, english is not my native language, so I may sometimes mix up words.

Perhaps we are kind of onto the same idea, when you talk about the changing of frequency sensitivity, it may be the same effect as the change in level sensitivity of a grouping of several single cones. A single cone can only adapt its intensity level sensitivity, not the frequency sensitivity, but combined with the same effect in other cones of other frequency sensitivity bands it creates a shift in the over all frequency sensitivity...

So, when you look at the picture, all the individual cones try to adapt to the level of light it is exposed to, this would cause that the cones that are exposed to the light from the same pixel of the image on the screen would, if regarded as a group, have a change in frequency sensitivity. This would be like to how the entire retina change sensitivity in a way that reduce the impact of the different frequency components of different lightsources.

[TheBlackCat]

Hmmm, I was not actually trying to say that the cone can not regenerate quickly enugh, the reduction is related to the cell trying to adapt to the level it is exposed to, you need less of the stuff when looking at something bright than something dark, I would think that regenerating so much of the proteins that you saturate the cells response would make you just as blind as loosing it all.
There is no protein that is regenerated. The protein itself is not what detects light, it is a small pigment molecule embedded in the protein. What happens is that when exposed to light, the molecule changes its shape and can no longer remain in the protein. It then triggers a cascade that ultimately causes ion channels to close, reducing the cell's activity electrical. It then travels out of the cell into a layer called the "pigment epithelium", which changes it back to its original shape. It then fits back inside the original protein, and the cycle repeats (assuming there is still light). What actually gets saturated in bright light is not the pigment molecule, it is the ion channels that the pigment molecule closes. The more channels are closed, the less channels are available to close when the next triggering of the cascade to take place. But that is not what is happening here, that sort of individual cell adaptation is always less than network adaptation involving a large number of cells, such as what I am discussing.

Perhaps we are kind of onto the same idea, when you talk about the changing of frequency sensitivity, it may be the same effect as the change in level sensitivity of a grouping of several single cones. A single cone can only adapt its intensity level sensitivity, not the frequency sensitivity, but combined with the same effect in other cones of other frequency sensitivity bands it creates a shift in the over all frequency sensitivity...[/quotes]
Cones sensetive to one frequency range feed back onto cones of another frequency range, enhancing or supressing the output of the cone (but not its level of activity, these can be two completely different things in cones). The cones themselves are not changing their behavior, what is changing is the electrical activity in horizontal cells, which communicate between different cones. Since horizontal cells sum the input from a large number of cones their impact on light sensetivity comes into effect well before individual cones get a chance to undergo any changes. That is "network adaption", as opposed to "receptor adaptation" which is what your at talking about. Computer models have shown that network adaptation is fully sufficient to account for the illusion seen here (it is a very popular illusion, so a lot of work has been done on it), receptor adaptation is not necessary.

[quote=TrAI] So, when you look at the picture, all the individual cones try to adapt to the level of light it is exposed to, this would cause that the cones that are exposed to the light from the same pixel of the image on the screen would, if regarded as a group, have a change in frequency sensitivity. This would be like to how the entire retina change sensitivity in a way that reduce the impact of the different frequency components of different lightsources.
Read my previous post about color opponency and simultaneous color contrast, I go into more detail there. Rods do not act independently. There is no signal travelling to your brain saying "this cone is active, this other cone is not". There is simply not enough space in the optic nerve to send that sort of signal, nor is particularly conducive to pattern-matching which is what the visual cortex is set up to do. What happens is that cones are combined into sets, with a center region surrounded by an opposing donut-shaped region. What is actually being sensed is the difference in light intensity or color between those two regions. That is the information that is actually sent to the brain. The actual comparison is largely done at the receptor level, with horizontal cells altering receptor synaptic activity based on the behavior of nearby cells. Ultimately, the signal that is being sent is merely what areas have changes in color or changes in contrast, not absolute levels of color or contrast (there are a handful of cells that do that, but there is no evidence they play any role here nor are they necessary to explain the phenomenon).

The issue is that, at least in color vision, these regions usually overlap. So what happens is that feedback between the horizontal cells and the receptors causes the receptors detecting a certain color to reduce the signal they send relative to their electrical activity than they normally would (so the same electrical activity would be percieved as less light). This also causes them to reduce their input back to horizontal cells, reducing the supression of opposing receptors and thus causing them to send a larger signal than they normall would. (remember a large signal is caused by less electrical activity and or less synaptic activity)

[TheBlackCat]

So it is as I said that it's the whitepoint compensation that's fooled.
Sort of, but not exactly. That is perceptually what is happening, but not on a cellular level. Practically no retinal cells can carry that sort of information. What is really being changed is the color contrast sensetivity for certain color pairs, and your brain ultimately interpolates between regions of color contrast in order to fill in regions of uniform color. Since it is happening with a large number of cells, it has the overall property of being percieved as a change in whitepoint, but that is not what the nervous system is actually doing.

[farmerjumperdon]

Did you know that only the central focal part of your vision is in colour, and the peripheral part is basically b&w? the brain fills in the colour info needed to give a seemingly fully coloured vision. and that's why it remains in colour until you move your eyes. The brain assumes the colours in the peripherical part to be constant as long as it has no new info. So until you move your eyes, you keep on seeing the colours there were before, giving the colour image when added to the B&W. Apparently there is some shift (even inversing) because it seems to be no simple addition of colour and B&W.

Perhaps the long exposure alters the sensitivity of your receptors, making them less sensitive for those colours and hence inversing the colour scheme when these colours are suddenly gone.

But anyway, the basis is in that the brain fills in the peripheral colour info of your vision because your eyes can't see colour in the periphery.

I remember reading about that now that you mention it. Rods & cones. One is for detecting color, or is more color sensitive. The other is the opposite. And the ones at the periphery, the B & W ones, are better at detecting motion. That's why picking up low light goodies and moving things happens better at the edges of vision.

Right?

[Doodler]

I remember reading about that now that you mention it. Rods & cones. One is for detecting color, or is more color sensitive. The other is the opposite. And the ones at the periphery, the B & W ones, are better at detecting motion. That's why picking up low light goodies and moving things happens better at the edges of vision.

Right?

I recall reading much the same.

[TheBlackCat]

I remember reading about that now that you mention it. Rods & cones. One is for detecting color, or is more color sensitive. The other is the opposite. And the ones at the periphery, the B & W ones, are better at detecting motion. That's why picking up low light goodies and moving things happens better at the edges of vision.

Right?

You are mixing up rods vs. cones with M ganglion cells vs. P ganglion cells. Rods and cones detect light, ganglion cells carry the information from the retina to the brain. Rods can only detect black and white, but they only operate at very low light levels. Even at twilight your rods are getting saturated. Cones send their input to both M ganglion cells and P ganglion cells. P ganglion cells carry color information and are only sensetive to light in a very small area of the retina. This means they are able to pick out fine details due to only looking at a small area, but are not as sensetive to motion or changes in light intensity. M cells are sensetive to light in a much larger area of the retina and are more sensetive to changes in light intensity and moving objects, but don't carry any color information (even though the cones the get their input from are color-sensetive). There are very few, although not zero, M ganglion cells in the center of the vision, but the color-sensetive P cells are scatterd throughout the retina.

[Lance]

Does anyone have a decent guess on what is being done to the color image? I have tried duplicating the illusion by inverting and over saturating but I can't seem to get the desired effect.

[TheBlackCat]

Invert works well enough for me. What program are you using?

[Lance]

Adobe PhotoShop v7.0

You may notice that if you take the image from the example in the OP (http://www.johnsadowski.com/2006_stuffs/manzana1.jpg) and just invert it, you do not restore the original. There is more going on.

[TheBlackCat]

That is because the "black" channel has been deleted. First, paste the original image into the photoshop or open it in photoshop. Invert it. Then go to Image-->Mode and change it to "CMYK color". Next, go to This Link (http://www.johnsadowski.com/2006_stuffs/manzana2.jpg) and copy that image (the image itself, not the link). Do not paste it yet. Then, got to Window and if Channels is not already checked, click it. Find the channels window on the screen. You should see CMYK, Cyan, Magenta, Yellow, and Black. Black should be pretty much empty. Click on "Black", then hit "paste". You will get the original image. Going to such lengths is not necessary, your eye will do the illusion just as well for the black parts of the image if you let it. If you really want to reproduce the effect exactly like they did, just open any image, set it to CMYK color, click on the Black channel, hit the "delete" button, then invert it. That will reproduce the effect exactly.

For a neat experiment, after you reproduce the image as I showed you. Try deleting the channels and then hitting "undo" to replace them one-at-a-time (don't delete the CMYK channel, that deletes everything). If you delte any of the color channels, it is still perfectly clear what you are looking at. It may be colored a little strangely, but it is still perfectly clear what it is. On the other hand, if you delete the black channel the image is practically unintelligable. If you look closely you can figure out what it is, but it is no where near as clear. This black is very important for your vision, your ability to interpret it is greatly reduced without it, and your ability to pick out fine details is pretty much gone. The color information really isn't that important. There are similar issues with sound, the exact waveform of the sound (the fine structure) is not really important, it is the envelope (the course structure) of the waveform that your brain really pays attention to when interpreting sounds (the fine structure is more important for telling where a sound is coming from). You can cut out the fine structure of one sound, put it under the envelope of the second sound, and you will interpet it as the second sound. The same goes for vision. Try fliping the corrected image (without the black channel) upside-down either through "rotate 180 degrees" or "flip vertically". Then paste the black-and-white image into the black channel like you did before. You will see a castle with slightly weird colors. The black information completely overrides the color information so much that you can scarcely tell the color information is actually an image at all and not just noise.

[turbo-1]

Lance, I think what's going on is that he took out the grayscale channel and then inverted the color information that was left. This way, when you mouse-over and get the grayscale image, the inverse after-image in your optical system "tints" the B&W image.

Oops, you got me TheBlackCat.

[Lance]

It seems I can more faithfully reproduce the effect by switching to the "Lab Color" mode, then inverting a combination of the A and B channels and overlaying the Lightness channel. This seems like it should be about the same thing as using the CYMK channels, but the inverted color image is different using the two methods.

[TrAI]

Does anyone have a decent guess on what is being done to the color image? I have tried duplicating the illusion by inverting and over saturating but I can't seem to get the desired effect.

The image seems to be the inverted chroma information of the original image, not just an oversaturated one. There are many ways to do something similar, this is one way that worked for me:

1: load the image in photoshop, use RGB mode
2: copy/paste to get three layers with the same image
3: desaturate the lowest layer
4: fill middle layer 50% gray
5: select top layer, set blending mode to color
6: merge top layer and middle layer
7: invert resultant layer
8: try the effect by watching this layer, then switch of layer visibility to display the desaturated version.
You can play with the the color layer to improve the effect, I used the "levels" function with input levels 90 1,00 140 to good effect, but this depends on the image.

Adobe PhotoShop v7.0

You may notice that if you take the image from the example in the OP (http://www.johnsadowski.com/2006_stuffs/manzana1.jpg) and just invert it, you do not restore the original. There is more going on.

Hmmm.. You could try to load the grayscale one, paste the color one to a layer, then invert this layer and lastly blend the two by using the color blend mode on the color layer. This would reproduce the original in the way I described previously in this post, unless you had changed the color layer, that is.

[Lance]

Here are two examples I just made. In both cases, the color image will show for 30 seconds, then display the swicth for 10 seconds.

The first was made by dropping the Luminance information and inverting the Chroma information. The second one was made by dropping the Black channel and inverting the CYM channels.

http://i58.photobucket.com/albums/g259/LlanceDaLlama/LlamasCL.gif

http://i58.photobucket.com/albums/g259/LlanceDaLlama/LlamasCYMK.gif

[TrAI]

There is no protein that is regenerated. The protein itself is not what detects light, it is a small pigment molecule embedded in the protein. What happens is that when exposed to light, the molecule changes its shape and can no longer remain in the protein. It then triggers a cascade that ultimately causes ion channels to close, reducing the cell's activity electrical. It then travels out of the cell into a layer called the "pigment epithelium", which changes it back to its original shape. It then fits back inside the original protein, and the cycle repeats (assuming there is still light). What actually gets saturated in bright light is not the pigment molecule, it is the ion channels that the pigment molecule closes. The more channels are closed, the less channels are available to close when the next triggering of the cascade to take place. But that is not what is happening here, that sort of individual cell adaptation is always less than network adaptation involving a large number of cells, such as what I am discussing.

Hmmm. Now that I think about it, my post may have been a bit badly worded, you are right its not the protein itself that detects light, however it is my understanding that it is the retinal that is bound in the photopsin that detects light, not the pigment. The pigments, the opsins, works kind of like filters, tuning the photopsin to a freqency band dependant on the opsin type. Opsins are proteins, so the pigment is a protein. If i recall correctly, the isomerization of retinal will trigger a change in the molecule, so that it will again trigger the transducin. Of course, I may have completly missunderstood the entire thing...

Cones sensetive to one frequency range feed back onto cones of another frequency range, enhancing or supressing the output of the cone (but not its level of activity, these can be two completely different things in cones). The cones themselves are not changing their behavior, what is changing is the electrical activity in horizontal cells, which communicate between different cones. Since horizontal cells sum the input from a large number of cones their impact on light sensetivity comes into effect well before individual cones get a chance to undergo any changes. That is "network adaption", as opposed to "receptor adaptation" which is what your at talking about. Computer models have shown that network adaptation is fully sufficient to account for the illusion seen here (it is a very popular illusion, so a lot of work has been done on it), receptor adaptation is not necessary.

Read my previous post about color opponency and simultaneous color contrast, I go into more detail there. Rods do not act independently. There is no signal travelling to your brain saying "this cone is active, this other cone is not". There is simply not enough space in the optic nerve to send that sort of signal, nor is particularly conducive to pattern-matching which is what the visual cortex is set up to do. What happens is that cones are combined into sets, with a center region surrounded by an opposing donut-shaped region. What is actually being sensed is the difference in light intensity or color between those two regions. That is the information that is actually sent to the brain. The actual comparison is largely done at the receptor level, with horizontal cells altering receptor synaptic activity based on the behavior of nearby cells. Ultimately, the signal that is being sent is merely what areas have changes in color or changes in contrast, not absolute levels of color or contrast (there are a handful of cells that do that, but there is no evidence they play any role here nor are they necessary to explain the phenomenon).

The issue is that, at least in color vision, these regions usually overlap. So what happens is that feedback between the horizontal cells and the receptors causes the receptors detecting a certain color to reduce the signal they send relative to their electrical activity than they normally would (so the same electrical activity would be percieved as less light). This also causes them to reduce their input back to horizontal cells, reducing the supression of opposing receptors and thus causing them to send a larger signal than they normall would. (remember a large signal is caused by less electrical activity and or less synaptic activity)


Hmmm, I see what you mean now. It does make sense, the differences in brightness of a picture on a screen is rather small compaired to the entire range of illumination that we can see. Is it maybe more on large differences, like between a dim place and a bright one(like direct sunlight) that one would expect what you called receptor adaption to play a bigger role?

[TheBlackCat]

Hmmm. Now that I think about it, my post may have been a bit badly worded, you are right its not the protein itself that detects light, however it is my understanding that it is the retinal that is bound in the photopsin that detects light, not the pigment. The pigments, the opsins, works kind of like filters, tuning the photopsin to a freqency band dependant on the opsin type. Opsins are proteins, so the pigment is a protein. If i recall correctly, the isomerization of retinal will trigger a change in the molecule, so that it will again trigger the transducin. Of course, I may have completly missunderstood the entire thing...
I'll have to check on this...




Hmmm, I see what you mean now. It does make sense, the differences in brightness of a picture on a screen is rather small compaired to the entire range of illumination that we can see. Is it maybe more on large differences, like between a dim place and a bright one(like direct sunlight) that one would expect what you called receptor adaption to play a bigger role?
Pretty much. The human visual system is set up to operate over approximately one log unit of light intensity. However, the entire visual range is about 10-12 log units. The visual system shifts that one log unit range as-need so that it is centered roughly on the mean illumination of the environemnt. Receptor adaptation and network adaptation both play a role in shifting the visual range. I think there is a limit as to how much change the network can deal with, but the receptors can deal with a very large degree of change. However, I think the receptors may be a bit slower, and both are far slower than the pupils (the pupils can only change light levels by a factor of six if I can remember correctly, as opposed to the factor of about 1000000000000 that the retina can accomodate).

[Nicolas]

So what is it now, can your periphery see color or not, less or more than the fovea? I'm lost in all proposals here :). The graphs I've seen indicate that the vast majority of color capable technology ;) is located in the fovea, with only the extremities of the "bell curve" in the peripheral part. So not none at all, but very little colour sensitive gear in the peripheral part.

That would explain why Argos saw the grey coming back at the center first.

[umop ap!sdn]

Why didn't I see this thread before... :o

The pigments, the opsins, works kind of like filters, tuning the photopsin to a freqency band dependant on the opsin type.
Yes, but not filters in the same sense as a photographer might use.

Retinal by itself absorbs a band of wavelengths (or frequencies) with peak absorption in the ultraviolet, somewhere around 320nm IIRC. Various factors contribute to the shift into the visible region: the opsins in our red and green cones include a chloride binding site, and there are several sites along the seven helices that surround the retinal where the presence or absence of a hydrophilic group has an electrostatic effect on the retinal molecule, shifting its wavelength. Here (http://www.genetics.org/cgi/content/full/158/4/1697) is a paper that describes the difference between the human red and green sensitive photopigments - merely 3 sites collectively account for the shift in peak wavelength. :D

Edit: in the human, it's 3 sites, not 5.

So what is it now, can your periphery see color or not, less or more than the fovea? I'm lost in all proposals here :).
Anecdotally, I once tried randomly lighting red and green LEDs seeing them only out of the corner of my eye, and could not tell the difference - they all looked yellow - but I could easily distinguish yellow from blue. So while there is some color perception (out of a sample of one :) ) it might be qualitatively reduced at the extremes.

[HenrikOlsen]

Since we seem to have several who knows about the chemistry of vision, has any studies been made on the variability of vision, eg. how large the difference can be from the average?

Most people I know lose color vision when it gets dark, which is why I normally have no problems going unseen in a forest at night while wearing a dark red cloak, but one of my friends apparently keeps his color perception even in very dark situation and can pick me out easily because he still see the color contrast from the background.

[Goblin]

That is a very cleaver trick using retinal after images. They happen in the retina and not the visual cortex. If it were in the cortex it would transfer to the opposite eye which it does not.

The spot is to make sure the image burns into area of the retina that will superimpose on the mouse over photo. That way when you move the mouse over the retinal image will be perfectly aligned with the picture on your cpu monitor and you get the cool color effect.

1. If you stare at different locations it will not work, the retina will not receive a constant image in the same location.

2. Close your left eye and stare at the dot for 30 seconds. When you mouse over the picture close the right eye and view with the left. It will be in black and white. If it were occuring in the visual cortex you would see color in the left eye.

He had to choose the proper colors to get the desired after image effect. A nice illusion for sure.








http://www.johnsadowski.com/big_spanish_castle.html#

[Goblin]

Your retina is quite capable of keeping the pigment level constant, at least for cones. That does not happen with cones, in fact your cones will be destroyed long before you get enough light intensity to seriously overwhelm the refilling mechanisms (assuming nothing disrupts retinal metabolism, the low pressure and thus low oxygen in an airliner is enough to dim vision somewhat).
.

You can overwhelm the cones without destroying them. Staring at light bulb at 1 foot for about 30 seconds will do the trick. When you look away you will be blind for a short period and then when regeneration occurs your vision will fade in good as new.

People with macular degeneration will take much longer to recover vs normals. Its called a photostress test. Before medications were available staring at a lightbulb was a treatment for glaucoma. The iris constricts which allows better flow of aqueous out of the eye which lowers IOP.

[Goblin]

I remember reading about that now that you mention it. Rods & cones. One is for detecting color, or is more color sensitive. The other is the opposite. And the ones at the periphery, the B & W ones, are better at detecting motion. That's why picking up low light goodies and moving things happens better at the edges of vision.

Right?

Another reason your peripheral retina is better in low light is the ratio of rods to ganglion cells. In the fovea its 1 to 1 for high resolution. The rods individulaly are more sensitive to light (1 photon is all that is needed) plus up to 200 rods can converge to one ganglion cell. This is why you only have 20/400 peripheral vision, very low rez but good at collecting photons over a larger surface area.

This is also why seeing dim stars is easier using eccentric fixation.

[Chunky]

i did it wrong at first i stared at the black and white image for a while and then i moved the mouse over the pic and when it changed colors it looked like it was spinning, it was odd...well not spinning but it tilted

[TheBlackCat]

Another reason your peripheral retina is better in low light is the ratio of rods to ganglion cells. In the fovea its 1 to 1 for high resolution. The rods individulaly are more sensitive to light (1 photon is all that is needed) plus up to 200 rods can converge to one ganglion cell. This is why you only have 20/400 peripheral vision, very low rez but good at collecting photons over a larger surface area.

This is also why seeing dim stars is easier using eccentric fixation.

There are no rods in the fovea. The fovea is all cones.

You can overwhelm the cones without destroying them. Staring at light bulb at 1 foot for about 30 seconds will do the trick. When you look away you will be blind for a short period and then when regeneration occurs your vision will fade in good as new.

This is not a matter of overwhelming the receptors, this is a matter of adapting to higher light intensities then having to re-adapt to lower intensities. That is different than what happens in the rods, which completely shut down.

That is a very cleaver trick using retinal after images.
See the previous page for a more in-depth discussion of simultaneous color contrast.

They happen in the retina and not the visual cortex. If it were in the cortex it would transfer to the opposite eye which it does not.
I've heard this many times, but it is not true. Many aspects of visual processing stay segregated by eye even at the cortex. They are mostly segregated in the primary visual cortex, with only a very small portion dealing with binocular processing. And seperation between the input from the two eyes remains, to varying degrees, through secondary visual cortex (V2), tertiary visual cortex (V3), and even into higher visual cortex areas like the MT. So it could easily be a cortical effect. But other evidence I discussed earlier and would prefer not to re-type a third time indicates it probably takes place in the earliest stages of retinal image processing.

[TheBlackCat]

Since we seem to have several who knows about the chemistry of vision, has any studies been made on the variability of vision, eg. how large the difference can be from the average?
There can be massive difference in the relative number and distribution of the various receptors. However, somehow human vision (except for people with defects of some sort) is remarkably uniform. There has to be some sort of compensatory mechanism that corrects for the disparity in retinal input, but we are not exactly clear on what it is or how it works.

[HenrikOlsen]

Could part of it be in the postprocessing?
We're likely the only animal who has a lot of communication about what we see, so part of our understanding is filtered by input from other people?

[Goblin]

Black cat I know the fovea is cones. I don't think I stated that but if I did it was a typo.

My point was rods are not a one to one connection like cones in the fovea are. Plus rods are more sensitive to photons vs cones.

You raise a good point on after images being cortical. The lack of ocular transfer infers retina but as you stated it could be monocular VC nerons. I have seen some cases that indicate this. Usually is someone with damage to V1 area. Some of these patients complained of persistant after images. I have seen one that did have ocular transfer to the opposite eye after having monocular stimulas. This suggests at least in some part there can be cortical . This transfer almost never occurs in healthy subjects.

It has be thought after images were due to photo pigment bleaching. But I do remember at least one study probably 5 years ago that found two types, one being local (retinal) and another appears to require cortical adaptation.

Quite true that not all the VC neurons are binocular. Why don't these neurons transfer after images to the opposite eye? One study suggested involvment of cortical suppression. Patients with damage to the visual cortex and report after images that do transfer support this theory.

While pigment bleaching certainly is the source for some after images (people with retinal desease have much longer after images) there is support for a cortical component as well.




There are no rods in the fovea. The fovea is all cones.

I've heard this many times, but it is not true. Many aspects of visual processing stay segregated by eye even at the cortex. They are mostly segregated in the primary visual cortex, with only a very small portion dealing with binocular processing. And seperation between the input from the two eyes remains, to varying degrees, through secondary visual cortex (V2), tertiary visual cortex (V3), and even into higher visual cortex areas like the MT. So it could easily be a cortical effect. But other evidence I discussed earlier and would prefer not to re-type a third time indicates it probably takes place in the earliest stages of retinal image processing.