Luminous Landscape Forum
Raw & Post Processing, Printing => Colour Management => Topic started by: fike on July 11, 2010, 08:48:43 am
-
I recently got an X-rite Color Checker passport. It seemed like a convenient little device to help me neutralize the color cast on my 10-stop ND (different story).
I messed with it a bit and put it away until I had a non-photographic application for it. I wanted to see if I could use my color managed workflow to help me match paint on my house so that they could scan a printed color at the hardware store.
I photographed the exterior wall in question with the color checker passport sitting in front of it. It was early evening on an overcast day--so there was even diffused light. I downloaded the raw files and converted to DNG before making the calibration file with the color checker software. I used Adobe ACR to apply the calibration and make a tiff. I then used my custom printer profile on my epson 2400 to print out the resulting color-corrected tiff.
The color wasn't even close.
I do use a color-calibrated monitor, but really, in this case, a monitor profile shouldn't make any difference. I did NO subjective manipulation of the image. I expected to get a decent approximation, but the wheat (off white) color wasn't really close enough to use for an approximation. Are my expectations too high, or is there something I am missing.
-
In this case, yes I think your expectations are a little high.
What you are trying to do is produce an accurate scene referred color in a print, but you are missing a step, because you haven't determined exactly what that color is.
For this, you need to measure the color of the paint on your exterior wall with a spectrophotometer and then process the file to reproduce that in print as close as possible.
Without the external reference, you are hoping that the sensor response and raw conversion produce an accurate color, which is then printed as it appears on screen. That is not likely (as you've found out) to be the case.
To produce an accurate scene referred color, you need to process to specifically produce that color. But without a reference of what that color is, the best you can do is trial and error until you obtain a print that is a close match.
We are relatively good at seeing hue differences in two references held up side by side, so by holding the print up against your paint, even slight differences will be visible.
The easiest way you'll produce a close (but still not exact) match will be to take a paint scraping to the paint supplier and have them measure it with a spectrophotometer. They'll then be able to produce something close.
Regards,
Peter
-
Another factor: Did you soft-proof the file? Softproofing will allow you to see a simulation of how the file will appear when printed. If you duplicate the file and use softproofing on the original, you can then see how to adjust the file to look as much like the "un-softproofed" version of it as possible.
-
In this case, yes I think your expectations are a little high.
What you are trying to do is produce an accurate scene referred color in a print, but you are missing a step, because you haven't determined exactly what that color is.
For this, you need to measure the color of the paint on your exterior wall with a spectrophotometer and then process the file to reproduce that in print as close as possible.
Without the external reference, you are hoping that the sensor response and raw conversion produce an accurate color, which is then printed as it appears on screen. That is not likely (as you've found out) to be the case.
To produce an accurate scene referred color, you need to process to specifically produce that color. But without a reference of what that color is, the best you can do is trial and error until you obtain a print that is a close match.
We are relatively good at seeing hue differences in two references held up side by side, so by holding the print up against your paint, even slight differences will be visible.
The easiest way you'll produce a close (but still not exact) match will be to take a paint scraping to the paint supplier and have them measure it with a spectrophotometer. They'll then be able to produce something close.
Regards,
Peter
Perhaps my expectations ARE too high, but I think you missed the fact that I calibrated with one of these: http://www.xritephoto.pth4.com/custom_page...+Rite+Photo+-+G (http://www.xritephoto.pth4.com/custom_page.aspx?pageid=141&utm_source=google&utm_medium=cpc&utm_term=colorchecker%20passport&utm_campaign=X+Rite+Photo+-+G)
-
Perhaps my expectations ARE too high, but I think you missed the fact that I calibrated with one of these: http://www.xritephoto.pth4.com/custom_page...+Rite+Photo+-+G (http://www.xritephoto.pth4.com/custom_page.aspx?pageid=141&utm_source=google&utm_medium=cpc&utm_term=colorchecker%20passport&utm_campaign=X+Rite+Photo+-+G)
No, I didn't miss the fact at all.
The answer still stays the same. Just because you produce a profile off an x-rite (I also use it), doesn't mean you are going to get an accurate scene referred color in a print.
Regards,
Peter
-
Perhaps my expectations ARE too high, but I think you missed the fact that I calibrated with one of these: http://www.xritephoto.pth4.com/custom_page...+Rite+Photo+-+G (http://www.xritephoto.pth4.com/custom_page.aspx?pageid=141&utm_source=google&utm_medium=cpc&utm_term=colorchecker%20passport&utm_campaign=X+Rite+Photo+-+G)
But I missed the term Absolute Colormetric in the color management. And no mention whether you created the profile at 6500K or whatever color temperature the picture was made at and no mention at what color temperature the two samples where compared. I would expect that the housepaint lost any FBA effect in time if it ever had any) but the paper could have FBA aboard. With all the conditions brought in balance you still may see a metameric failure as the inks used on the print build a color with another and less opaque layer structure and most likely another set of pigments. And the Color Checker isn't an absolute calibration tool either for the same basic reason.
Yes, your expectations are too high. On the other hand they bring back some down to earth reality to the entire color management discussion. I see a lot of absolute statements on CM in the forums while in practice I know it can be very difficult to reproduce colors correctly of a painting or even a color drawing in a print. A print of a landscape photographed several weeks ago and 5000 miles away has a lot of conditions less strict but also gives more freedom on what color is representative for that scene.
You are right about the monitor. I shouldn't play a role in this test.
The practical side has a flaw too. The new paint color mixed to the weathered paint will have a shift in color in less than a month after it is applied. Beter check whether there is an old paint can in the shed of the paint used and mix the new paint to that leftover paint sample in the can. Painters here use a RAL swatchbook and paint mixing recipes + in some cases a spectrometer that can measure wet and dry paint samples. Often that method can still show differences after the paint is used as layer thickness and surface texture play a role too.
met vriendelijke groeten, Ernst Dinkla
Try: http://groups.yahoo.com/group/Wide_Inkjet_Printers/ (http://groups.yahoo.com/group/Wide_Inkjet_Printers/)
-
hmmm...interesting. So if I had a tool to measure color temperature, or estimated it more accurately, this procedure might have worked better. I see your point about the brightness of the white in the paper, but they were custom profiles so I would assume that would be somewhat accounted-for.
If I were to use the 17% gray card in the color checker, I might be able to get the color temp a bit closer. I acknowledge that this won't ever be perfect, but it seems like a good exercise of my workflow, minus the actual buying and applying paint.
So, from a purely experimental point of view, what should I try to get as close as possible?
just for our edification, here is a jpg of what I did, I know the colors going to sRGB will be all goofed-up from their original aRGB.
(http://www.trailpixie.net/temp-image/home-color-check-50D-7086-2-2.jpg)
-
hmmm...interesting. So if I had a tool to measure color temperature, or estimated it more accurately, this procedure might have worked better. I see your point about the brightness of the white in the paper, but they were custom profiles so I would assume that would be somewhat accounted-for.
If I were to use the 17% gray card in the color checker, I might be able to get the color temp a bit closer. I acknowledge that this won't ever be perfect, but it seems like a good exercise of my workflow, minus the actual buying and applying paint.
So, from a purely experimental point of view, what should I try to get as close as possible?
I could sum up more possible flaws in that method. What you want to achieve would have some practical use in reproduction photography if the painting and the reproduction are viewed in the same environment later on, which is not often the case. The gamut of the painting should then also fit within the printer's gamut. Where a print doesn't have the gamut and dynamic range of outdoor scenes this experiment has little practical use anyhow as you probably might get closer to the post's off-white paint but your neighbour's car color clipped at the boundary of the printer's gamut. I do not see how this Color Checker>Camera>Profile creation>Sample print ever could replace a spectrometer reading of the paint or much better a simple paint swatchbook metameric match to the original paint with your eyes as the observer, twice done, at noon and in the early morning. Nothing will beat the last system. Your wife may have even better eyes than you have so ask her to confirm your choice. In the end you two will be most frequent observers of that freshly painted post so there will not be a metameric failure due to another observer :-)
CM in (reproduction) photography has its place but often enough I make that print proof and edit the image to get the best representation of the original next to it.
met vriendelijke groeten, Ernst Dinkla
spectral plots of +100 inkjet papers:
http://www.pigment-print.com/spectralplots/spectrumviz_1.htm (http://www.pigment-print.com/spectralplots/spectrumviz_1.htm)
-
No, I didn't miss the fact at all.
The answer still stays the same. Just because you produce a profile off an x-rite (I also use it), doesn't mean you are going to get an accurate scene referred color in a print.
I’m in agreement with Peter here. The tool was used for something its not designed for. Same with the entire camera system used with this extra tool.
I messed with it a bit and put it away until I had a non-photographic application for it.
That’s kind of the issue.
-
I guess I was surprised by how far off it was. Is the color temp measurement of the shooting environment the primary issue? I think I understand well issues of paper profiling, and I am very familiar with normal color variance between paper types and manufacturers on custom profiled papers. I have done test printing on dozens of different papers with far less variance than this.
-
I guess I was surprised by how far off it was. Is the color temp measurement of the shooting environment the primary issue? I think I understand well issues of paper profiling, and I am very familiar with normal color variance between paper types and manufacturers on custom profiled papers. I have done test printing on dozens of different papers with far less variance than this.
First things first. You shot a raw right, not a JPEG (as indicated above)?
-
First things first. You shot a raw right, not a JPEG (as indicated above)?
Of course.
-
I guess I was surprised by how far off it was. Is the color temp measurement of the shooting environment the primary issue? I think I understand well issues of paper profiling, and I am very familiar with normal color variance between paper types and manufacturers on custom profiled papers. I have done test printing on dozens of different papers with far less variance than this.
I would think that the color temp of the shooting environment may have some to do with it, if most definitely not all.
One thing I have found with using white balance swatches is that they just make everything seem neutral. For photos taken under typical natural light somewhere around mid-day that is fine enough, otherwise, simple balancing to a single white balance patch makes a mess IMO. Try doing that for a sunset photo or a golden evening lighting photo or a tungsten photo or some cold-blue winter evening light, etc. Once you apply that correction they look nice an neutral sure, but NOTHING like they did in real life and many times it can totally destroy the photo.
I wonder if making a dual-light color checker camera calibration and then setting camera white balance by the white patch reading would allow it to work more naturally. I really need to get around to making a dual-light camera calibration for ACR.
-
I messed with it a bit and put it away until I had a non-photographic application for it.
Color photographs can do some odd things to human perception that fine art landscape painters have grappled with for centuries. From that posted image your camera's perception of overcast to me seems quite blue. Not sure if that paint really looks that way under that version of overcast. Camera's AutoWB tend to render overcast quite blue, but in my area of the country, central Texas, overcast makes objects appear warmer. See image sample below.
An isolated color patch or paint swatch can be any slight variation of the desired color but still look close in a photo on account of what the surrounding interplay of cool and warm colors in a photograph do to the perception of these patches. In color science it's called color constancy:
http://en.wikipedia.org/wiki/Color_constancy (http://en.wikipedia.org/wiki/Color_constancy)
This is why it takes a machine (spectro) having a fixed and constant light source to define the color of that paint to eliminate the effects of color constancy and varying color temp appearances.
Also some pigment brands and paint bases can reflect back unexpected hues after the paint has dried when viewed under different light sources. I had this problem painting a still life with tempura paint. It was rich looking when wet and went chalky when dried. After it dried, one rich color of blue looked great under soft white incandescent and went purple next to window daylight.
In this image note how the appearance of neutral changes depending on the surrounding hue of colors. What looks neutral isn't always R=G=B.
[attachment=23128:ColorPat...tPhotoLL.jpg]
-
(http://www.trailpixie.net/temp-image/home-color-check-50D-7086-2-2.jpg)
Fike - Here is your jpeg with the Color Checker colors more in the ballpark. Is this now a better match to the paint color?
[attachment=23145:inCamera...7086_2_2.jpg]
Cliff
-
Fike - Here is your jpeg with the Color Checker colors more in the ballpark. Is this now a better match to the paint color?
[attachment=23145:inCamera...7086_2_2.jpg]
Cliff
Yes! That is very close. What did you do?
PS the RAW police are going to have to take away your color management license because you did it with a jpg file.
-
Yes! That is very close. What did you do?
PS the RAW police are going to have to take away your color management license because you did it with a jpg file.
By correcting the Color Checker patches to match target values (I used the actual target values of my old 24 patch CC), the corrections for the CC also corrected the paint color. Most colors can be corrected this way, unless the spectrum is really weird.
I used Pictographics inColor to make a profile. I don't know what went wrong with the profile you made with the Color Checker software - I haven't tried it but I expect there is a way to get all the CC patches to match target values. Maybe all the ACR settings have to be neutral when you apply it?
-
By correcting the Color Checker patches to match target values (I used the actual target values of my old 24 patch CC), the corrections for the CC also corrected the paint color. Most colors can be corrected this way, unless the spectrum is really weird.
I used Pictographics inColor to make a profile. I don't know what went wrong with the profile you made with the Color Checker software - I haven't tried it but I expect there is a way to get all the CC patches to match target values. Maybe all the ACR settings have to be neutral when you apply it?
Wow! Crames! I just checked the Lab readouts on several patches on your version of Fike's image and you nailed just about every dang one of them.
It's just now the overall color cast appearance of the rest of the scene looks way too greenish beige. Now I'll admit from my posted motorcycle image that overcast appearances differ in other parts of the world, but this looks like the kind of color cast often associated with the sun low in the sky right after a rain shower. At least it looks this way in Texas.
Fike, did the overall cast look this way at the time you took the shot or was it more neutral looking?
So the paint on your house IS actually yellowish looking. Your comment about the print looking (wheat/off white) which I associated with looking too yellow threw me off as to exactly how the paint color actually appears.
Interesting to see another piece of software produce different results compared to what I get using Adobe's free DNG Profile Editor to generate a camera profile.
-
but this looks like the kind of color cast often associated with the sun low in the sky right after a rain shower. At least it looks this way in Texas.
Fike, did the overall cast look this way at the time you took the shot or was it more neutral looking?
BINGO!
Yes, that is what the paint looks like. I didn't expect a perfect match, but your sample is MUCH better.
Now, what did I do wrong? When I purchased this little doohickey I expected that I would get results like this by putting a set of known good colors in the scene and then calibrating against them. It appears that this isn't what happened with the software. As you may know, the ColorChecker software doesn't have much in the way of controls to modify what it is doing.
Where should I start? I see that you used some third party software. Are there any other possible options besides buying another piece of specialized software?
-
BINGO!
Yes, that is what the paint looks like. I didn't expect a perfect match, but your sample is MUCH better.
Now, what did I do wrong? When I purchased this little doohickey I expected that I would get results like this by putting a set of known good colors in the scene and then calibrating against them. It appears that this isn't what happened with the software. As you may know, the ColorChecker software doesn't have much in the way of controls to modify what it is doing.
Where should I start? I see that you used some third party software. Are there any other possible options besides buying another piece of specialized software?
(I guess you were asking me although you quoted tlooknbill)
I haven't tried the Color Checker software but I will take a look at it when I have a chance. Maybe Andrew, madmanchan, or someone else familiar with it can offer guidance? Another alternative might be the Color Checker calibration technique of Bruce Fraser, or the actions derived from it by Simon Tindemans and others. I don't have the links handy, but they were popular a few years ago and should be easy to find.
But I would be surprised if the Color Checker software can't do the same thing.
Cliff
-
Where should I start?
Calibration and color management with ICC profiles are two different processes that must come together for optimal results. In this case you've made a good profile but haven't calibrated the device for the lighting you were shooting in. You might say we "calibrate" cameras by selecting the optimal exposure and white balance for every lighting situation we work with. In this case, the image appears to be underexposed and the white balance is too cool. This topic can get very deep, especially when we start talking about making prints that match the original objects, but I think that the jist of what you're looking for.
-
Calibration and color management with ICC profiles are two different processes that must come together for optimal results. In this case you've made a good profile but haven't calibrated the device for the lighting you were shooting in. You might say we "calibrate" cameras by selecting the optimal exposure and white balance for every lighting situation we work with. In this case, the image appears to be underexposed and the white balance is too cool. This topic can get very deep, especially when we start talking about making prints that match the original objects, but I think that the jist of what you're looking for.
I think you are missing what has occurred. I photographed a known standard target. I tried calibrating to that target and got poor results. Cliff used different software, working on my JPG, and got results that were subjectively-speaking much closer. I am not looking for a primer on color management. I am looking for ideas on what could have gone wrong with the Passport color checker software.
-
I tried calibrating to that target and got poor results.... I am looking for ideas on what could have gone wrong with the Passport color checker software.
I'm quite familiar with the process. You didn't calibrate, you've *profiled* the process and implemented it *without calibration* (proper exposure and white balance). Proper exposure and white balance are what's missing.
-
I'm quite familiar with the process. You didn't calibrate, you've *profiled* the process and implemented it *without calibration* (proper exposure and white balance). Proper exposure and white balance are what's missing.
Then how did Cliff (Crames) get such good results? I thought that the purpose of the color checker was to help me set the white balance and make a good profile for a unique lighting condition.
-
Then how did Cliff (Crames) get such good results?
He's applied color corrections to compensate for the lack of proper white balance and exposure. It's kinda like taking a poor film scan and making it look good. These days we can avoid color correction by setting our white balance and exposure correctly with digital cameras.
I thought that the purpose of the color checker was to help me set the white balance and make a good profile for a unique lighting condition.
Yes, the white balance of the image you attached, however, wasn't set correctly and neither was the exposure. That's were you need to focus your efforts - white balance and exposure.
-
Then how did Cliff (Crames) get such good results? I thought that the purpose of the color checker was to help me set the white balance and make a good profile for a unique lighting condition.
The inCamera software that I used generates an ICC profile from the ColorChecker that corrects not only hues and chromas, but also makes automatic changes that affect the tone curves and white balance. It forces all of the patches to match target values exactly in lightness, chroma, and hue. Other colors in the scene then fall into place.
I think the difference is that dng profiles don't adjust the lightness/contrast or white balance. You set those manually, and then the camera profile gets the various hues and chromas in line. As far as I know, it doesn't matter if you use the ColorChecker Passport-generated profile, or create one with the Adobe DNG profile editor, it's the way camera profiles work.
Like Onsight said, to use a dng camera profile in ACR, not only do you assign the custom camera profile, but you also have to click a neutral patch for white balance, and adjust exposure to a reasonable setting (for example set exposure so that the lightest gray patch has a green channel of 143 if your target working space is AdobeRGB). Set all of the other setting like "recovery", "fill light", "blacks", "brightness", etc. to zero. This will get you closer to getting the target values of the patches on the Color Checker and other colors in the scene.
Cliff
-
Did you set the white balance using the neutral gray on the color checker? Make sure you select the correct one
-
Did you set the white balance using the neutral gray on the color checker? Make sure you select the correct one
I think that is probably the major root of my problem. I expected the color-checker profiling software would have done that for me. I haven't had a chance to get back to them, but I'll do that when I have some time for learning and experimentation.
-
Yes I think you find that things will fall into place when the white balance and exposure are set correctly. Start by clicking on the Passport's white patch with the white balance tool and adjust the Exposure to achieve the values that math that of a CC reference file. Afterwards, work on achieving the ideal exposure in camera so as to minimize development steps. These are your *calibration* steps that will compliment the *profile* that you've generated. Again, calibration and profiling are two different processes that work together.
-
I think that is probably the major root of my problem. I expected the color-checker profiling software would have done that for me. I haven't had a chance to get back to them, but I'll do that when I have some time for learning and experimentation.
The DNG profiles are not meant to correct for white balance, if they did they would only be useful for that specific color temperature (and even then only when you wanted a neutral WB, which for creative reasons is not always the case). So when applying your profile to an image you still need to set the appropriate WB setting in raw converter (which should be easy in your case since you can just use the eye dropper on one of the neutral patches in the CC checker, generally the second lightest patch is considered the best bet).