Reds

[privatebydesign]

A very large portion of jpeg compression comes from the fact that human vision is notoriously insensitive to colour. That is why compression algorithms convert RAW sensor data to the YCbCr colour space, to separate out luminance (Y) from colour (CbCr), the human eye is much more sensitive to detail, luminosity variation, than colour and the colour component can be compressed much more than the luminosity and still retain more information than the human eye can discern at normal viewing distances.

 

[Dick]

I have faced the problem with reds, unfortunately.

I'm having serious problems with reds these days. My daughters often have red clothes on and the photos taken require lots of PP to fix the red channel issues. Working with sliders, curves, levels, ... what ever Aperture offers, is just not enough, but the adjustments then need to be brushed on the red bits only. For example lowering saturation to a point where a red jacket starts to look ok, already turns skintones to crap.

There is no real solution as far as I know. I have tried to build presets with no luck. In some cases I have been forced to turn photos into BW, because fixing the red channels has been too difficult. Now I'm starting to think that it is not really possible to take photos of red objects. Just like I don't go out in bright sunlight to take photos, I maybe shouldn't take photos of red things.

 

[jrista]

something else to think of, we are unable to remember colors. Our brain is just no wired that way. There are some exceptions to the rule, but they are few and far between. . . I work with colors in my daily job and have learned that my memory tricks me often and now have Pantone books in strategic locations.

pierre

I'd like to see the scientific papers that explain how and why this could be true. I remember colors quite clearly myself for a good while. It is only after years as memories naturally fade, and our recollection of all details, not just color, diminishes. There have been a few papers written on the subject of memory and color, and the science demonstrates that we actually remember things better when they are in color, particularly NATURAL colors, vs. when they lack color, or are artificially colored.

In that context, I can easily understand the OP seeing the red as his images are being replicated by LR, and not remembering the color of the real flowers that way. The real flowers obviously have natural color, whereas the rendition of those flowers in LR is less natural. That discrepancy, and the OP's sensation of it, more than anything else here, is absolutely real. The closer the OP can get his flowers to render like the real natural color, the less he will feel that something is off.

There are limitations with rendering realistic color when it comes to digital photography. For one, silicon is extremely sensitive to red and infrared light. To attenuate this sensitivity, camera manufacturers use IR cutoff filters that have a gradual incline in the passband as it enters the reds. This limits how well we can naturally resolve reddish colors. Further, most computer screens are 8-bit sRGB, and the sRGB gamut is a rather limited one. It does not allow the greatest extent for deeper reds. Combine that with potentially limited or incorrect recording of reds by the camera sensor, especially for brighter colors that may be close to the non-linear shoulder of the sensor's response (i.e. closer to the clipping point of the signal), and getting realistic reds can be very difficult.



Calibration with a color checker card is one way to improve results. I wouldn't call it a panacea, but it can help when your working with extreme colors. Keeping your exposure in the linear range of the sensor will also help. That usually means reducing the exposure a little bit, which might hurt shadow detail if you have any...but it will preserve more accurate colors at the high end of the signal.

Manual tweaking of colors with the color channel sliders in LR can do a lot, but it works best if all of your channels are sufficiently below the clipping point of the signal. You need some headroom to shift colors around without unnatural casts appearing.

Sorry, no papers, but here's what I have read and what seems to be going at my work.
As explained by martti, our vision system is relative rather than absolute. When we try to memorize the colors, we mentally compare them to something we know or describe them as a warm red, darker than something else and so on. We are creating a reference point that can be used for later recognition.
try this, have somebody make you 10 squares of various red color. pull one out of the hat and mark the back. Look at the color and try to memorize it without rationalizing its hue, saturation or brightness. Put it back in the hat and have the person helping you lay them out on the table next day. See if you can pick the right one. If you concentrate on the color alone without making mental references to it being dark, light, warm, cold and so on, it is very unlikely that you'd be able to pick the right square.

let me know how it goes!

pierre

I'm not sure why you would try to remember a color without remembering the way you would describe it. Our memory is about associations and contexts...to try and remember something out of context or the associations we make with it is an exercise in futility. I was never talking about spectrum or the eyes sensitivity to any particular part of the spectrum or how we mathematically define colors or how we name them or anything like that.

I was referring to the way people remember color. We remember color the way we expect color, and in that respect, remember colors best when they seem appropriate. In that context, it is easy to look at a photo with red flowers in Lightroom, and feel the colors are wrong.

This isn't some colorimetric thing, it's a psychological, perceptual thing. The CIE Lab space maps the range of colors visible to the human eye, and most computer color gamuts do not reach as far into the reds, violets, or blues (particularly, if the later versions of CIE's Lab space that utilized a 10 degree foveal spot in testing are any indication) as our vision allows us to see.

I have a ton of tulips planted in my yard, spanning the color range. Whenever I take a photograph of them, I can never get the reds or purples as rich and deep...yet still as detailed...as the flowers are in real life. The information burns out and loses microcontrast if I try to push the color saturation for any given channel near the reds or violets or magentas that far, and the saturated colors are still rarely right. Not only do I know the colors are wrong, I can step outside and verify they are wrong.

It's not the spectrum, nor is it the specific "color" as mathematically modeled that we remember. We remember color as we have come to expect it. Natural colors are easiest to remember, and bright, vibrant, saturated colors that we come across in nature are usually the first we sense are wrong in our photographs. At least, I rarely ever feel that I've captured the non-green, non-earth colors well in my photos.

 

[mb66energy]

I sometimes struggle to get reds to look "right" out of my 6D and 5D MkIII. Attached is a recent picture of some bright red tulips taken with my 6D and 135L lens. I tried different profiles in Lightroom and played with red and green hue, saturation and luminescence, but the red color doesn't match the color I remember seeing. I ended up using the neutral profile, which comes closest to my recollection. I have not tried setting up my own profiles yet.

Anybody else experiencing these problems? Any suggestions on how to correct?

Thanks

I added some info to a thread several months before because I "suffered" from the limited capability in reds:
http://www.canonrumors.com/forum/index.php?topic=22204.msg424412#msg424412

@jrista: Perhaps green is easier to reproduce because red and blue channels help to adjust the green. But red and blue have to live on it's own - there is no IR or UV channel supporting color reproduction from the "left" or the "right" side of the red/blue pixels spectrum.

Funny: After getting two 5D classic cameras the problem is more or less solved. They give very good color reproduction for reds (Poppies, red tulips), much better than 40D, 600D or EOS M.

Some remarks to things I read here:

• Someone stated that the CRI in a forest is 1.0. Never. CRI = 1.0 describes an ideal black body spectrum provided (mostly) by the sun, incandescent bulbs, candles.
  Introducing any filtering reduces parts of the spectrum or enhances a more or less narrow band of the spectrum. A forest shows more green than red and blue in the spectrum. Look at the forest scene. Changes are subtle but this tiny correction gives a much more shades of grey green.
  
• 
  * The "right color" is maybe very sensitive to the perception of different individuals but the color differences within one object will be recogniced by most of the individuals similarily (hopefully it is still english language .
  The two poppy flower images show the differences in the rendering of the texture of the leaves - just here it's subtle but changing the K-value from 5200 -> 4400 gives more structure in the reds and helps to discriminate them.
  This image is made with a 5D classic which gives me good results without postprocessing - mentioned above.
  ]

Comparison between no color corrections

 

[danski0224]

In line with the 5D comment above-

There is something different about the color rendering from older Canon cameras.

The 1D, 1Ds-1DsIII and 5D (Mk I) all have positive comments specific to color rendering.

Coincidentally, none of those cameras have good high ISO performance.

I have read posts from others in the past about the sensor CFA actually altering colors. One post over on FM illustrated how some colors reproduced in one camera but not another.

Maybe the push to have cameras with an ISO range of 100-26000 has taken color away.

I can say with certainty that files from a 1DsIII are distinctly different from a 5D3.

Maybe the 5Ds/r are attempts to correct the vanishing colors. If Canon makes a worthy 1DsIV, I'm in.

 

[mb66energy]

In line with the 5D comment above-

There is something different about the color rendering from older Canon cameras.

The 1D, 1Ds-1DsIII and 5D (Mk I) all have positive comments specific to color rendering.

Coincidentally, none of those cameras have good high ISO performance.

I have read posts from others in the past about the sensor CFA actually altering colors. One post over on FM illustrated how some colors reproduced in one camera but not another.

[...]

Good to hear another ones perception of the color rendering of the 5D classic (and it's "siblings").

Perhaps it's a matter of narrower filter spectra cutting out more light but giving a stronger definition for red, green and blue channels. The trade of is having less light on the photo sensors which leads to a lower ISO sensitivity ...

Oh this world with never ending compromises ... but I like the 5D classic very well. It is one of the better compromises at least for my photography.

 

[danski0224]

While I do not own a 5D mk1, the differences in colors between the camera bodies that I have tried are noticeable.

I actually like the 1D (4.2mp) output, but there is no room for cropping. Maybe part of it is the CCD sensor. I should take it out again soon.

The other issue with that body is no AFMA, and no Canon service. No way to tune the lens to the camera.

 

[bluemoon]

something else to think of, we are unable to remember colors. Our brain is just no wired that way. There are some exceptions to the rule, but they are few and far between. . . I work with colors in my daily job and have learned that my memory tricks me often and now have Pantone books in strategic locations.

pierre

I'd like to see the scientific papers that explain how and why this could be true. I remember colors quite clearly myself for a good while. It is only after years as memories naturally fade, and our recollection of all details, not just color, diminishes. There have been a few papers written on the subject of memory and color, and the science demonstrates that we actually remember things better when they are in color, particularly NATURAL colors, vs. when they lack color, or are artificially colored.

In that context, I can easily understand the OP seeing the red as his images are being replicated by LR, and not remembering the color of the real flowers that way. The real flowers obviously have natural color, whereas the rendition of those flowers in LR is less natural. That discrepancy, and the OP's sensation of it, more than anything else here, is absolutely real. The closer the OP can get his flowers to render like the real natural color, the less he will feel that something is off.

There are limitations with rendering realistic color when it comes to digital photography. For one, silicon is extremely sensitive to red and infrared light. To attenuate this sensitivity, camera manufacturers use IR cutoff filters that have a gradual incline in the passband as it enters the reds. This limits how well we can naturally resolve reddish colors. Further, most computer screens are 8-bit sRGB, and the sRGB gamut is a rather limited one. It does not allow the greatest extent for deeper reds. Combine that with potentially limited or incorrect recording of reds by the camera sensor, especially for brighter colors that may be close to the non-linear shoulder of the sensor's response (i.e. closer to the clipping point of the signal), and getting realistic reds can be very difficult.



Calibration with a color checker card is one way to improve results. I wouldn't call it a panacea, but it can help when your working with extreme colors. Keeping your exposure in the linear range of the sensor will also help. That usually means reducing the exposure a little bit, which might hurt shadow detail if you have any...but it will preserve more accurate colors at the high end of the signal.

Manual tweaking of colors with the color channel sliders in LR can do a lot, but it works best if all of your channels are sufficiently below the clipping point of the signal. You need some headroom to shift colors around without unnatural casts appearing.

Sorry, no papers, but here's what I have read and what seems to be going at my work.
As explained by martti, our vision system is relative rather than absolute. When we try to memorize the colors, we mentally compare them to something we know or describe them as a warm red, darker than something else and so on. We are creating a reference point that can be used for later recognition.
try this, have somebody make you 10 squares of various red color. pull one out of the hat and mark the back. Look at the color and try to memorize it without rationalizing its hue, saturation or brightness. Put it back in the hat and have the person helping you lay them out on the table next day. See if you can pick the right one. If you concentrate on the color alone without making mental references to it being dark, light, warm, cold and so on, it is very unlikely that you'd be able to pick the right square.

let me know how it goes!

pierre

I'm not sure why you would try to remember a color without remembering the way you would describe it. Our memory is about associations and contexts...to try and remember something out of context or the associations we make with it is an exercise in futility. I was never talking about spectrum or the eyes sensitivity to any particular part of the spectrum or how we mathematically define colors or how we name them or anything like that.

I was referring to the way people remember color. We remember color the way we expect color, and in that respect, remember colors best when they seem appropriate. In that context, it is easy to look at a photo with red flowers in Lightroom, and feel the colors are wrong.

This isn't some colorimetric thing, it's a psychological, perceptual thing. The CIE Lab space maps the range of colors visible to the human eye, and most computer color gamuts do not reach as far into the reds, violets, or blues (particularly, if the later versions of CIE's Lab space that utilized a 10 degree foveal spot in testing are any indication) as our vision allows us to see.

I have a ton of tulips planted in my yard, spanning the color range. Whenever I take a photograph of them, I can never get the reds or purples as rich and deep...yet still as detailed...as the flowers are in real life. The information burns out and loses microcontrast if I try to push the color saturation for any given channel near the reds or violets or magentas that far, and the saturated colors are still rarely right. Not only do I know the colors are wrong, I can step outside and verify they are wrong.

It's not the spectrum, nor is it the specific "color" as mathematically modeled that we remember. We remember color as we have come to expect it. Natural colors are easiest to remember, and bright, vibrant, saturated colors that we come across in nature are usually the first we sense are wrong in our photographs. At least, I rarely ever feel that I've captured the non-green, non-earth colors well in my photos.

OK, that's a valid point, but it does not invalidate my point that we have issues with remembering the color and that such a phenomenon could be the cause for the perceived problem under some circumstances.

pierre

 

[martti]

Re: Reds, the Polaroid guy Land was a genius, nothing less!

Now, please spend a while to appreciate the genius of Dr. Land.
Nobody in the realm of psychology had ever observed these phenomena.
They did not exist.
But his guy run the necessary scientific tests to show that this is one of the ways our sensory system creates a mapping or an image pattern that our brain can take advantage of in plannning approacing-retreating, fear or recognition..
No, he did not study the emotional connections, it was to fall on other cross-cultural scientists such as Christopher Koch..

Seriously creative guys! Hats off.