Ad
Ad
Ad
Pages: [1]   Bottom of Page
Print
Author Topic: Why do colors vary so much based on light temp?  (Read 6635 times)
Redcrown
Full Member
***
Offline Offline

Posts: 157


« on: February 21, 2013, 04:10:27 PM »
ReplyReply

Why do colors render much differently under different light temps, even when white balance is correctly set in ACR?

This has always been a mystery to me, and recent shots of an orchid gave a good example.

I shot the orchid under my soft box, first with just the incandescent modeling light, then with the strobe light. Then I moved it to a window and shot there (heavy overcast mid-day sky).

For each image, I shot a full-frame gray card image first and set a custom in-camera white balance. Then I included the graycard in these test images and re-set the white balance in ACR. Shots are from a Canon 5D2, the Adobe "Camera Neutral" profile was used in ACR.



The strobe light (4300K) and the window light (5400K) images are obviously very close in color. But the modeling light (2550K) image is significantly different. No white balance adjustments will make these images match.
Logged
Tim Lookingbill
Sr. Member
****
Offline Offline

Posts: 1264



WWW
« Reply #1 on: February 21, 2013, 04:50:00 PM »
ReplyReply

"Camera Neutral" profile most likely is a single 6500K illuminant expecting a close to neutral light source. See what you get with a custom DNG profile using a single 2800K illuminant table profile or create a dual illuminant one shooting two CCchart shots each lit under both illuminant light sources respectively.

Other than that, it may be your camera's IR/UV sensor filtering is being thrown off by the heavily red biased 2550K modeling light.
Logged
Jim Kasson
Sr. Member
****
Offline Offline

Posts: 1079



WWW
« Reply #2 on: February 21, 2013, 06:26:03 PM »
ReplyReply

Why do colors render much differently under different light temps, even when white balance is correctly set in ACR?

I can think of three possible answers.

1) I believe that ACR/LR uses simple coefficient scaling for WB adjustments. There are more sophisticated approaches, like von Kries (which uses coefficient scaling, but with the cone cell primaries). These might produce more accurate results in your example.

2) It is possible, even likely, that the illuminant metamerism that a human would observe and the illuminant metamerism that the camera observes are different. Here's what I mean by that. Case A: Colors that a human would see as matching under one illuminant might not match under one of the other illuminants; colors that a human would see as different under one illuminant might look the same under one of the other illuminants. Case B: Colors that the camera sees as matching under one illuminant might not match under one of the other illuminants; colors that the camera sees as different under one illuminant might yield the same RGB triplet under one of the other illuminants. It is unlikely that the colors in case A will be the same as the colors in case B.

3) Human color constancy is more complicated than matching individual color patches. Indeed, it is not fully understood.

Jim
« Last Edit: February 21, 2013, 06:29:45 PM by Jim Kasson » Logged

Schewe
Sr. Member
****
Offline Offline

Posts: 5544


WWW
« Reply #3 on: February 21, 2013, 09:47:53 PM »
ReplyReply

Why do colors render much differently under different light temps, even when white balance is correctly set in ACR?

Because unlike the eye which has color consistency (the ability to adapt to different colors of light) a sensor can not. What you are seeing is most likely due to a metameric failure of the sensor where the same color color (the flower) reflects light and is seen by the sensor as a different color under different spectral illumination. Daylight has a lot of blue, tungsten has a lot of red (and little blue). So, under different lights, the colors actually have different spectral reflectance curves.

The strobe and skylight appear very close (close enough to fix in WB adjustment with a tiny touch of HSL if needed) while the tungsten is a very different color and can not really be fixed with simple white balance.

It's a fact of life that nothing can really capture colors the exact same way human eyes can...that's where the art comes in...what color do you WANT the flowers to have?

Tip: the clue here is to not say "the way my eyes see them" because that ain't gonna happen.
Logged
MarkM
Sr. Member
****
Offline Offline

Posts: 336



WWW
« Reply #4 on: February 23, 2013, 06:09:26 PM »
ReplyReply

Blue flowers have been problematic for photographers from the beginning. Getting accurate images was a big problem even with film. Flowers that looked blue ended up reproducing with way too much red on slide film. The reason for this is that many flowers have strong reflectance in the far red end of the spectrum right before infrared. While the color is technically visible to us, our eyes have little sensitivity to this range. When you move your subject into tungsten light which is much stronger in the longer wavelengths the flower reflects a lot more of this light than will be apparent to your eyes. 

Here's a paper by Kodak from the good old days: http://www.kodak.com/global/en/professional/support/techPubs/e73/e73.pdf
Logged

Jim Kasson
Sr. Member
****
Offline Offline

Posts: 1079



WWW
« Reply #5 on: February 24, 2013, 10:17:42 AM »
ReplyReply

Mark, thanks for the link to the Kodak paper. It brought up a memory:

I was talking to Ed Giorgianni, a Kodak R&D guru, at one of the image-related conferences in the early 90s (SPIE, SigGraph, I don't remember). We got onto the fact that CCDs were so profoundly sensitive in the infrared, and what a problem it was to get the right filtration. He told me that it was a problem in film, too. A wedding photographer had once sent him a picture of the husband-to-be, the best man, and all the ushers. The image had been made on color negative film.  All were wearing black tuxedos. None of the tuxedos was black in the picture. None of the tuxedos was the same color as any one of the other tuxedos. Ed said that the black aniline dyes used to dye clothing are strongly reflective in the infrared, and that the three layers of the film had some sensitivity there. The result was a disaster for the photographer.

I had a Leica M8 with similar problems. The solution was the same as in the film days: UV- or IR-cut filters over the lens. Leica, corporately red-faced, supplied them for M8 buyers, but only one free one to a customer.

Jim
« Last Edit: February 24, 2013, 10:20:20 AM by Jim Kasson » Logged

thierrylegros396
Sr. Member
****
Offline Offline

Posts: 728


« Reply #6 on: February 24, 2013, 02:08:05 PM »
ReplyReply

Thanks for all answerers of this Topic !

It explains so much things.

For examples why some Blue-Magenta flowers were almost impossible to be reproduced even on a well calibrated screen !
And appear very different when shooted in natural shadow or with fill flash.

And also why some fluorescent tubes in trains look red-orange in reality and more Red-Blue when reproduced on screen.

I thought metamerism was only related to paper and ink reproduction.

Have a Nice Day.

Thierry

Logged
Redcrown
Full Member
***
Offline Offline

Posts: 157


« Reply #7 on: February 25, 2013, 01:33:54 PM »
ReplyReply

My thanks, too, for the responses. They led me to a lot more reading on the physics of light. Probably a lot more info than I'm willing or able to absorb.

My question was mostly an academic one, just to satisfy a curiosity. I'm rarely in search of "true" color, and farily adept at achieving "pleasing" color. And for what it's worth, that orchid came with a warning by the florist that the colors were artificial. Apparently achieved by feeding it colored water. The warning said that when it re-blooms it will be white.
Logged
Jim Kasson
Sr. Member
****
Offline Offline

Posts: 1079



WWW
« Reply #8 on: February 25, 2013, 01:44:12 PM »
ReplyReply

...that orchid came with a warning by the florist that the colors were artificial...

Try shooting it through an IR-cut filter and see if it's like the tuxedos.

Jim
Logged

Jim Kasson
Sr. Member
****
Offline Offline

Posts: 1079



WWW
« Reply #9 on: February 25, 2013, 01:50:06 PM »
ReplyReply

My thanks, too, for the responses. They led me to a lot more reading on the physics of light.

If you're not ready for a lot of linear algebra, you could do a lot worse than this book: http://www.amazon.com/Digital-Color-Management-Wiley--Technology/dp/047051244X/ref=sr_1_2?s=books&ie=UTF8&qid=1361821565&sr=1-2&keywords=digital+color+management
Logged

RFPhotography
Guest
« Reply #10 on: February 25, 2013, 02:20:31 PM »
ReplyReply

FWIW, my thoughts are similar to Jeff's.  Colour constancy and metameric failure.  Even in human vision it falls down at the extremes of the light temperature scale.  That's why we can 'see' the warmth of sunrise/sunset and the blue of shade.  The grey card is quite close based on the custom WB.  But the orchid reflects such a narrow part of the spectrum that it's significantly impacted by the equally narrow part of the spectrum that's lighting it.

Logged
Redcrown
Full Member
***
Offline Offline

Posts: 157


« Reply #11 on: February 25, 2013, 02:38:37 PM »
ReplyReply

OK, another question arises...

What's the difference between a $150 to $200 "IR Cut filter" and the IR filter that is supposedly built in to the sensor on most digital cameras?
Logged
madmanchan
Sr. Member
****
Offline Offline

Posts: 2110


« Reply #12 on: February 26, 2013, 12:18:23 PM »
ReplyReply

Blue/purple flowers are particularly hard for cameras to render correctly.  As noted above this has to do with the near-IR response of these flower petals and the IR cutoff on the sensor.  Camera profiles optimized for these types of flowers will often mess up on color accuracy for other more common types of materials, and similarly profiles optimized for common objects tend to struggle with these types of flowers. 
Logged

Jim Kasson
Sr. Member
****
Offline Offline

Posts: 1079



WWW
« Reply #13 on: February 26, 2013, 02:49:14 PM »
ReplyReply

What's the difference between a $150 to $200 "IR Cut filter" and the IR filter that is supposedly built in to the sensor on most digital cameras?

It's impossible to say without more information, but my first guess would be (since the filter sounds expensive and you probably didn't mean that it's expensive because it's for a 400mm f/2.8 lens) that the screw-on IR-cut filter is of the interference, aka dichroic, aka thin-film, type, whereas the one in the camera is absorptive. The interference filters are more complicated (ergo more expensive) to make and can usually produce steeper skirts (transitions between the pass and stop band) than available in absorptive filters.

That's just a guess. Filter manufacturers will tell you a lot about their products, but camera companies are as a rule less forthcoming.

Jim
Logged

Ellis Vener
Sr. Member
****
Offline Offline

Posts: 1863



WWW
« Reply #14 on: February 27, 2013, 07:04:16 PM »
ReplyReply

My thanks, too, for the responses. They led me to a lot more reading on the physics of light. Probably a lot more info than I'm willing or able to absorb.

My question was mostly an academic one, just to satisfy a curiosity. I'm rarely in search of "true" color, and farily adept at achieving "pleasing" color. And for what it's worth, that orchid came with a warning by the florist that the colors were artificial. Apparently achieved by feeding it colored water. The warning said that when it re-blooms it will be white.

Try "QED: The Strange Theory of Light and Matter" by Richard Feynmann: http://www.amazon.com/QED-Strange-Theory-Light-Matter/dp/0691024170/ref=sr_1_1?ie=UTF8&qid=1362013414&sr=8-1&keywords=QED+Feynmann
Logged

Ellis Vener
http://www.ellisvener.com
Creating photographs for advertising, corporate and industrial clients since 1984.
Schewe
Sr. Member
****
Offline Offline

Posts: 5544


WWW
« Reply #15 on: February 27, 2013, 08:42:01 PM »
ReplyReply

Try "QED: The Strange Theory of Light and Matter" by Richard Feynmann

Actually, your link went to the previous edition...there a new addition available/
Logged
orchidblooms
Full Member
***
Offline Offline

Posts: 160


« Reply #16 on: March 02, 2013, 08:43:27 PM »
ReplyReply

i have been struggling for years with orchids...

invested in olympus e-1 and the good ed lenses for their good colors - i finally got colors on orchids acceptable except for shades of plum/pink  and lavender / purple

now i am using d800e and zeiss lenses.

my rendering of colors has improved greatly but still not where i would like them  

my home depot setup is an array of 6-8 simple reflectors with ge daylight cfls... i put parachute fabric over them - use 2 silver/ gold reflectors and a few makeup mirrors to try to get the color/light for what i am looking forr...

seems i am ready for strobes...

or redoing my HD array with photo cfls...

looking at a profoto d1 kit

also looking / leaning twords... a broncolor minicom 80 kit...

generally i like to invest in things only once - i used the e-1 for years and years...  in fact i was finally ready do move to the e-5 last march when the d800s came out...  changed things for me....

anyone have thoughts specifically on strobes for floral portraiture?

huge thanks

Phil

« Last Edit: March 02, 2013, 09:42:00 PM by orchidblooms » Logged
Jim Kasson
Sr. Member
****
Offline Offline

Posts: 1079



WWW
« Reply #17 on: March 03, 2013, 02:37:42 PM »
ReplyReply

i have been struggling for years with orchids...my home depot setup is an array of 6-8 simple reflectors with ge daylight cfls... i put parachute fabric over them - use 2 silver/ gold reflectors and a few makeup mirrors to try to get the color/light for what i am looking for...

Phil,

It's hard for me to offer anything definitive because I don't know the details of your lighting/camera color filter array, but I can give you a few things to think about. The first has been mentioned already: IR-cut filters. The second has not received any explicit attention in this thread: the smoothness of the spectra of your lights. Compact fluorescent lamps usually offer emission spectra that are quite peaky. There are versions that moderate these tendencies; they usually advertise a high Color Rendering Index (CRI). I would advise trying some CFL's with CRIs of around 100. There's no guarantee that they will work better for you than what you've got now, but it could be a cheap experiment compared to buying new strobes.

There are several reasons why a CRI of about 100 is not a dispositive condition for better color rendering in photography. The first is that the CRI is predicated on the human cone color sensitivities, not the sensitivities of the filters in your particular camera. The second is, as the Wikipedia article referenced above states, the CRI is not a perfect index of the effect it is trying to measure. Still, A CRI around 100 tends to mean a smooth spectrum, and a smooth spectrum light source tends to be better for avoiding unexpected color shifts. I would avoid sources with CRIs much above 100, because they may use tricks to emphasize portions of the spectrum to fool the eye into thinking colors are richer than they otherwise might be perceived, and those tricks might not work on your color filter array, and might nt produce results you like even if they did.

You say you are using colored reflectors to modify the light. If the source has a bunch of narrow spectral peaks, the reflected source will (usually) have them too, but the relative amplitude will be affected. Therefore, you can't filter a peaky light source into one with a smooth spectrum.

In general, it would be a good thing for you to look at the spectra, or at least the CRI, of any light source you're considering. (strobe, HID, LED, etc). An incandescent source will naturally have a high CRI if there are no tricks employed in the design, but they tend to run to warm color (photographically warm, which, confusingly, means a cooler color temperature), which may not be what you want for orchids. They're also hot, which is almost certainly not what you want for orchids.

LEDs are not usually stellar CRI performers, but they are a thermally cool source, and are readily available with color temperatures of 5000 degrees Kelvin and higher. There is a series of LED bulbs produced by Cree that, while not intended for photography, have CRIs in the mid 90s. Here's one, but it's 2700 deg K.

Hope some of that helps.

Jim
Logged

orchidblooms
Full Member
***
Offline Offline

Posts: 160


« Reply #18 on: March 03, 2013, 02:46:48 PM »
ReplyReply

jim...  huge thanks

i think i will try greater wattage cfl bulb that is promoted for photog work...

i will check the ratings as you suggest...  both of our local pro camera shops offer these cfl bulbs...

in the long run - the strobes will allow me to do more / shoot color gels against black backgrounds...

and allow opportunities i think to better capture the texture of flowers... (fast crisp...)

but 90.00 compared to - for starters 2300.00 is sure worth a look

huge thanks
Logged
Jim Kasson
Sr. Member
****
Offline Offline

Posts: 1079



WWW
« Reply #19 on: March 03, 2013, 06:32:02 PM »
ReplyReply

Pleased to help, Phil.

While pitched at lighting for print viewing, this article offers some interesting information about D50-ish light sources -- note the peaks in the fluorescent spectrum. A D50 task lamp is an interesting idea for your orchids. I use Macbeth D50 fluorescents in my studio, and find them close, but not the same as, a D50 incandescent source. I've never tried serious photography under them.

Jim
Logged

Pages: [1]   Top of Page
Print
Jump to:  

Ad
Ad
Ad