How Dogs See Things : Another Challenge!

Jonathan Neal wrote on 4/20/2007, 1:13 PM

A Photo of Dog Vision

I was thinking that this might be a fun project, and maybe especially for those video editors who have kids.

Man's Best Friend has dichromatic vision -- meaning that Dogs can see only part of the range of colors in the visual spectrum of light wavelengths. Simply put, dogs are red-green color blind. This means that they see in shades of yellow and blue primarily. They see a brighter and less detailed world when compared to humans.

Humans, on the other hand, have trichomatic vision, meaning that they can see the whole spectrum.

More specifically for today's challenge, most humans see the rainbow of colors described by "VIBGYOR": Violet, Indigo, Blue, Green, Yellow, Orange, and Red (plus hundreds of variations on these shades) while dogs see "VIBYYYR" (Violet, Indigo, Blue, Yellow, Yellow, Yellow, and Red). The colors Green, Yellow, and Orange all look alike to dogs; but look different from Red and different from the various Blues and Purples. Dogs are very good at telling different shades of VIB apart. Finally, Blue-Green looks White to dogs.

I was wondering if some Channel Blend or Color Correcting experts out there wanted to share the calculations that might be required to convert 'human-see' footage to something roughly like what our canine friends observe.

The photo I placed at the top was a rough guesstimate by Mark Plonsky, Ph.D. of what a dog and human might see when viewing a color band.

Comments

Tech Diver wrote on 4/20/2007, 2:31 PM

I'm not a biologist but as far as I know, nobody real knows how dogs perceive the colors that they do see. For example, what we see as yellow may be a totally different experience for a dog. It may be a "color" that is unlike anything we have ever seen and therefore indescribable. Merely removing green from the "normal" scene is how a human might see through dogs eyes but not necessary how a dog sees through them. Therefore my personal view is that you can choose to do anything with your image and no one will ever know if it is correct or not.

I have also heard that dogs are very good at detecting moving objects but are not too good at recognizing non-moving ones and they rely very much on the sense of smell to suplement this.Their brains may be applying an image differencing algorithm rather than pattern matching.

craftech wrote on 4/20/2007, 2:38 PM

While they do not recognize as many colors as we do, dogs recognize many more shades of gray than we can and as a result can differentiate objects that are composed of colors they don't see exactly the same as we do, but appear as subtle shades of gray they can recognize. There was an interesting ABC special on this a few years ago. They also said that the reason for the recognition of the various subtle gray shades was because they can see much better in dim light than we can. I remember them mentioning that a dog's field of vision is at a much greater angle - something like 260 degrees compared to a human's 180 degree field of vision. In other words, they can see around themselves much better than we can.
Interesting.

John

dand9959 wrote on 4/20/2007, 2:46 PM

As a lifelong dog owner, I suspect that regardless of color, everything a dog sees looks like food or a toilet.

PeterWright wrote on 4/20/2007, 6:58 PM

My dog's quite a good weather forecaster - I asked him how it's going to be tomorrow and he said "ruff".

vicmilt wrote on 4/20/2007, 8:02 PM

yeah - mine's a sports fan...
I asked her who was the single greatest player in baseball and she said, "Roof"

fldave wrote on 4/20/2007, 8:32 PM

"everything a dog sees looks like food or a toilet."

dand9959: thanks for making me spew beer all over my formally beautiful ViewSonic LCD monitor.

Thanks, I needed a laugh like that!

Owner of two large dogs...

Interesting thread, though. Thanks for thinking out of the box, Jonathan.

dsf wrote on 4/20/2007, 8:53 PM

>>>Tech Diver 4/20/2007 3:31:54 PM: “I'm not a biologist but as far as I know, nobody real[ly] knows how dogs perceive the colors that they do see.”

Finally, as having an honours degree in Experimental Psychology, I may actually have a clue about which I speak: Of course, no one can look through a dog’s eyes and see what he sees. If I were doing the experiment, I would show color illustrations of dots with different color combinations. Depending on the color combinations certain symbols would be visible. The dog would be trained to do something (e.g., go to the food chute) if he saw the experimental symbol (e.g., a circle). The perceptabillity of such a symbol would depend on his ability to differentiate different colors. And in that way it could be determined which colors he could differentiate. (This is exactly how the color vision of humans is determined.)

Now if I only knew how Vegas works.

Tech Diver wrote on 4/21/2007, 2:30 PM

Thanks for the insight DSF. As a dog owner (mini Dachshund) I would be interested in seeing the video you generate.

Peter

UKAndrewC wrote on 4/21/2007, 3:04 PM

Given the colour translation you have described, this is an approximation using the secondary colour corrector.

Start with the Desaturate All But Reds preset:

Under Chrominance set:

gain: 1.6

under limit hue, set:

center: 220
Width: 70
Smooth: 75

This wlll give you roughly what you outlined, reasonable Red Blue response with Green in monchorome. If you use this on the Gradient test pattern under media generators, it gives a good estimation of the colour response.

Of course the model you gave can't be correct as it gives a centre of strong Yellow and without a Green component you can't have a strong Yellow, this too would have to tend towards Grey.

Unfortunately the behavioral test wouldn't be conclusive. It could only prove that a dog can distinguish between hues and saturations, not colours.

It is however, possible to determine how a dog receives colour, based on the biological composition of the rods and cones in their eyes.

P.S. A quiz based on this topic, what is it that my Father can be, that Mother never can?

TTFN

Jonathan Neal wrote on 4/21/2007, 5:10 PM

Very nice.

I tried these settings and achieved something close to what the color pattern suggests:

Channel Blend



Color Space: RGBA



Red = 1.000 * R + 1.500 * G + -0.250 * B + -0.500 * A + 0.000

Green = 1.000 * R + 1.500 * G + -0.250 * B + -0.500 * A + 0.000

Blue = 0.250 * R + 0.000 * G + 1.000 * B + -0.500 * A + 0.000

Alpha = 1.000 * R + 1.000 * G + 1.000 * B + 1.000 * A + 9.000

Dog Color 1

Dog Color 2

Dog Color 3

dsf wrote on 4/21/2007, 10:11 PM

>>>Tech Diver: 4/21/2007 3:30:33 PM: “As a dog owner (mini Dachshund) I would be interested in seeing the video you generate.”
Hmm…I presume you have your tongue firmly in your cheek.

>>>UKAndrewC : 4/21/2007 4:04:54 PM: “Unfortunately the behavioral test wouldn't be conclusive. It could only prove that a dog can distinguish between hues and saturations, not colors.”

In the experiment I crudely described, all colors would be indistinguishable from one another to a color-blind subject; i.e., the hue and saturation would all be the same and the dots could only be distinguished by color vision: all dots would appear to be the same to a “color blind” subject. The dog could only distinguish the experimental pattern if he had the ability to distinguish the colors presented in the experiment; which colors would be varied to determine what colors he could differentiate.

>>>”It is however, possible to determine how a dog receives color, based on the biological composition of the rods and cones in their eyes.”

I agree that such a physiological examination of a dog’s eyes would be very useful. But there is no way to determine what colors the dog “sees” (i.e., in a psychological, perceptual sense) other than by an experimental procedure similar to what I have described.

UKAndrewC wrote on 4/22/2007, 3:40 AM

Hi dsf

I'm just being a pedantic purist. Because colour is subjective and colour space and perception of light frequencies is infinite.

All we can safely say is that a dog can or cannot percieve the particular colours used in the experiment.

But I'll shut up now ;-)

dsf wrote on 4/22/2007, 10:06 PM

UKAndrewC 4/22/2007 4:40:42 AM: "All we can safely say is that a dog can or cannot percieve the particular colours used in the experiment."

That is correct. But the experiment could be refined to determine very precisely what shades or hues the dog could perceive. You need a good knowlege of statistical analysis to do this as not all dogs respond the same; you would also need a lot of dogs.

Doing things like this is how experimental psychologists pay their mortgages.