Serene Life

by garik

submit your photo


Hall of Fame
View past winners from this year

Please participate in Meta
and help us grow.

Take the 2-minute tour ×
Photography Stack Exchange is a question and answer site for professional, enthusiast and amateur photographers. It's 100% free, no registration required.

It's well known that the cones in retina are unable to distinguish colors in dark, leaving us only the monochrome softer vision provided by rods. And personal experience tells this "switch" to monochrome is rather a continuous decline than an on-off event.

Provided with correct exposure, a camera doesn't care about light levels and records colors the same way as daytime. I suspect this is a major reason why it's hard to obviously express dusk or dawn (as could be observed in this question) - the photo will be looking very similar to a daytime one.

So perhaps photos of dusk/dawn could be post processed to make their resemblance to eye-witnessing closer. I'm interested in whether the sensitivity is lost to all colors synchronously (so simple desaturation should do the trick), or do some colors disappear from our cognition before others?

share|improve this question

2 Answers 2

up vote 5 down vote accepted

I used to teach this stuff at the university but it's long ago. Let's see if I can remember:

The brain does some interpretation of the raw signal. We have 3 types of cones (B, G, and Yellow) and then the lowlight rods (bluish-green). So you see already now that we process that to see reds, and it also shows how easy we can get defects that alter our colour perception, and also why (in additive colour spces) we can trick our brain to seeing yellow by sending pure red and green. The red/yellow cones also react a bit to the low end of blue, which is why we perceive purple as red and blue.

The blue cones are the most sensitive but then we have only 2% of them, so the resolution sucks. Green cones are 32% and they are medium sensitive. Red/yellow have the best resolution with 46% and they are least sensitive. That leaves a total sensitivity that is most sensitive to green-yellow, which is why the monochrome conversion that averages (R+G+B)/3 is a bad conversion, while the V in YUV is weighted such that yellow is brightest, hereafter green.

However the peak of the rods are lower in wavelength, on the bluer side of green with red completely dark. I guess this is why movies depict night using blue hues. So if you want to desaturate to simulate our night vision, you need to change the weights in the 3x3 matrix that computes saturation and "lightness".

This actually sounds like a fun project, so I might play around with this in Image View Plus More when I get the time. Thanks for planting this seed in my brain :)

OK, I played around with it now :)

I found the wavelengths of R, G, and B LCD monitors and correlated them with the sensitivity for luminosity and for our darkvision cones.

Luminosity chart for normal vision and dark vision

That's the way you get the weights for luminocity convert to greyscale (R*0.299 G*0.587 B*0.114). Similarly I find weights for the dark vision and make the grescale image based on these values. Blue is about 4 times more sensitive (Compensation for the loss of blue sky and extra sensitivity for the blue moonlight to see dangerous water?), green is 85% as sensitive (we are developed for living in mother nature), and red is 8% as sensitive. Similarly, I can adjust R G and B for a colour image to change the colours from a colour image into another colour image. By the assumption of "expectations" bringing back the colour into the perception, I "blend" with the original colours.

From Cone to Rod perception:

1|5
2|6
3|7
4|8
  1. Original
  2. Cone based Luminosity
  3. Rod based Luminosity (vegetation is untouched, and the contrast between skin and veggie is enhanced? And the dark jacket got enhanced!)
  4. Original with 50% brightness
  5. Cone to Rod adjusted blend 25%
  6. Cone to Rod adjusted blend 50%
  7. Cone to Rod adjusted blend 75%
  8. Cone to Rod adjusted image

One thing to notice is that image 8 is very blue. We should note that the white balance in the light is also much different when the blue sky is gone, and before the blue moon light comes in, and here I only show how the image would look if we could see the individual colours with the sensitivity curve of the rods.

From Cone to Rod perception

If I instead simulate our ability to white-balance I remove some blue, and convert to the rod based luminosity, and blend 50-50, it looks like this:

Less blue, fading greyscale

share|improve this answer
2  
It's apparently called the Purkinje effect. Agree that it would be an interesting project :) –  j-g-faustus Dec 6 '12 at 15:52

I don't know whether colour detection decreases equally for all colours, that's straying a little bit into the field of human biology!

However I do know that our monochrome night vision is totally different to looking at a greyscale image. I believe the brain knows the colours are really there and so suppresses the fact that you can't see them from conscious part of the brain.

The only way to display twilight images and have them appear correctly would be to photograph and process them using the full colour gamut, and then display them in very low light conditions so the brain does the same processing that it would with twilight images.

Finally I disagree that it's difficult to express dusk/dawn in a photograph - the colour of ambient light can shift significantly as the light fades, regardless of the brain's ability to see these colours, so a strong colour shift in addition to lower brightness levels should do the trick. The problem with the Reuters photograph in this question was more to do with the photographer correcting for the low levels of light.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.