Luminous Landscape Forum
Raw & Post Processing, Printing => Colour Management => Topic started by: stargazer on January 15, 2019, 06:12:52 am
-
Hello,
I successfully made a printer profile as I encapsulated the Spyder measurements data in a .ti3 Argyll file. I use Spyder hardware and software for measurements and Argyll CMS for profile generation.
The result seems quite good to me as I can measure on the print almost the same values as what I see on the screen (as LAB values).
However I found some limitations. Because the input values are not spectrum values but LAB, I can generate a profile only for 5000K (I read on the internet that this is the ColorTemp. of the Spyder light source).
I'd like to generate a profile with different color temperature for different ambient light. As spectral data is missing, on theory it should be possible to change it by math calculations.
So, is it possible to generate a profile with different color temperature from LAB values? (using Argyll CMS)
The other limitation I think is not from LAB data but rather from Perceptual intent generation. I found the generated Relative intent quite good but the Perceptual intent is kind of washed out and looks a little bit purplish. I know that it is kind of expected behavior but I find it is too different from the Relative.
Is there any way to generate a Perceptual intent with a little bit more colors and contrast? (or if any tweaking is possible before or during profile generation)
Many thanks in advance
-
You might be better off sending this inquiry to the argyll mailing list although Mr. Gill does frequent this forum.
-
I'd like to generate a profile with different color temperature for different ambient light. As spectral data is missing, on theory it should be possible to change it by math calculations.
I'm not sure what you mean. Print profiles are always based on reflectance, so naturally any print is the product of the illuminant and the print. i.e. Using relative colorimetric white corresponds to unprinted paper, and unprinted paper will be (very close to) the color of the illuminant.
Creating a profile from the spectral reflectance allows for the spectral interaction of the intended illuminant and the print, but (when using a relative colorimetric intent) white is still white - i.e. the color of the illuminant used to view the print on the unprinted paper.
The other limitation I think is not from LAB data but rather from Perceptual intent generation. I found the generated Relative intent quite good but the Perceptual intent is kind of washed out and looks a little bit purplish.
Is there any way to generate a Perceptual intent with a little bit more colors and contrast? (or if any tweaking is possible before or during profile generation)
The effect of Perceptual gamut mapping depends primarily on the source gamut you specified. Ideally it is the gamut of the images you are converting, or the gamut of the colorspace they are encoded in, if the images fill that colorspace.
-
Many thanks for the answer. I'd like to add some details to my question to make it more clear.
As far as I understand, Spyder Print spectrocolorimeter uses light source with ColTemp 5000K (I read it on the internet).
The output measurement values, which I export from SpyderPrint SW are in LAB color space.
So, the resulting color profile created with Argyll CMS from these LAB values is optimized for light source with ColTemp 5000K.
I don't see any way to generate a color profile optimized for different light source (with different CT). As I understand the reason for that limitation is that the input data is in LAB color space (spectral data is missing). I'd like to generate a color profile optimized for light source with CT 3500K.
So, one possibility is to convert my input LAB values. Is it possible to do some kind of conversion and to generate another color profile as I have LAB values as input data?
-
Many thanks for the answer. I'd like to add some details to my question to make it more clear.
As far as I understand, Spyder Print spectrocolorimeter uses light source with ColTemp 5000K (I read it on the internet).
The output measurement values, which I export from SpyderPrint SW are in LAB color space.
So, the resulting color profile created with Argyll CMS from these LAB values is optimized for light source with ColTemp 5000K.
Spyder Print's spectrocolorimeter does not use a CT5000 light source. It uses a set of colored LEDs. Measures, reflectance of each colored LED and scales it against their reference white patch. From this limited set of measurements it estimates Lab values, (specified relative to D50 XYZ) for each patch read. How well it works depends on the spectral reflectance of the printed ink. I haven't run across any actual tests on specific performance but the algorithms could be tuned for typical CYM ink sets much like scanners, which have only 3 channels (RGB) can be tuned to a specific printer. How well this is done is not spec'ed.
I don't see any way to generate a color profile optimized for different light source (with different CT). As I understand the reason for that limitation is that the input data is in LAB color space (spectral data is missing). I'd like to generate a color profile optimized for light source with CT 3500K.
You need to specify what you mean by "optimized" for 3500K. Let's say you get a Spyder reading of Lab=(50,0,0), a middle gray, what do you mean by optimizing that for 3500K?
So, one possibility is to convert my input LAB values. Is it possible to do some kind of conversion and to generate another color profile as I have LAB values as input data?
-
You need to specify what you mean by "optimized" for 3500K. Let's say you get a Spyder reading of Lab=(50,0,0), a middle gray, what do you mean by optimizing that for 3500K?
It seems like I have misunderstood something.
I have a color profile created from Spyder measurement data and Argyll CMS.
In my opinion when I print a picture the colors will look natural (close to what I see on my calibrated screen) if I observe it under light source with CT 5000K. If I put the same print under light with CT 3500K, the colors will look warmer than what I have on my screen.
If I have a color profile "optimized" for 3500K, then the colors of the printed picture will look natural under light source with CT 3500K and will look colder under light source with 5000K.
Is it correct?
-
It seems like I have misunderstood something.
I have a color profile created from Spyder measurement data and Argyll CMS.
In my opinion when I print a picture the colors will look natural (close to what I see on my calibrated screen) if I observe it under light source with CT 5000K. If I put the same print under light with CT 3500K, the colors will look warmer than what I have on my screen.
Right. So for example if your screen is set up for 5000K and you view an image of a Colorchecker on the screen and compare it to an actual or printed Colorchecker viewed with 5000K they should all agree. But if you view the Colorchecker actual or print under 3500K they will appear warmer than the screen image.
And what you want to do is create a print from a profile that lets you view the print under 3500K such that it matches the image on the monitor rather than matching a Colorchecker illuminated with 3500K
It isn't possible. This can't be done because the level of blue required to create a "white" 5000K from a 3500K illuminant exceeds 100% reflectance as the shorter wavelengths are not greatly attenuated.
If I have a color profile "optimized" for 3500K, then the colors of the printed picture will look natural under light source with CT 3500K and will look colder under light source with 5000K.
Is it correct?
-
And what you want to do is create a print from a profile that lets you view the print under 3500K such that it matches the image on the monitor rather than matching a Colorchecker illuminated with 3500K
That's a quite good example. Thanks.
It isn't possible. This can't be done because the level of blue required to create a "white" 5000K from a 3500K illuminant exceeds 100% reflectance as the shorter wavelengths are not greatly attenuated.
I think I understand now. Actually the "color set" is different under 3500K.
So if I want to make a picture to look more natural (closer to my screen) under 3500K, should I tweak it before printing and make it
colder or there is a more elegant solution of that issue?
-
I think I understand now. Actually the "color set" is different under 3500K.
So if I want to make a picture to look more natural (closer to my screen) under 3500K, should I tweak it before printing and make it colder or there is a more elegant solution of that issue?
A way to approximate what you want is to, in Adobe RGB use Photoshop curves and decrease the R from 255 to 206 and the G down to 244*. This will darken and shift the monitor image to the blue. Print it. Make sure you reset the monitor image to compare to the print under 3500K.
This keeps the blue as high as possible and drops the green and especially the red such that the print will look similar under 3500K light. However, it will be darker so you should move the 3500K lamp closer to match the overall luminance. This does what is called the wrong von Kries transform. It will get you close but, as you can see, the print will come out darker so you will need more light to reach the same luminance as the monitor.
* This was edited because the original values were for a gamma=1 version of Adobe RGB. For gamma=1, drop Red from 255 to 160 and Green from 255 to 232. The values in the preceding paragraph have been corrected for gamma=2.2 (standard Adobe RGB (1998))
-
I'm not sure I saw the answer I had in mind so please forgive me if I repeat a post from someone else.
The spectra of light gets involved in the print profiling process in two places:
1. The spectra of the light used to illuminate the print and take a measurement
2. The spectra of light used to convert spectral measurements to Lab for profiling purposes.
So, in the case of the Spyder device, it is illuminated by a light source (#1) which may or may not be anywhere near D50. The device also only outputs Lab values.
Due to these limitations you will not be able to "roll a profile" relative to a different light source.
If you were using a spectrophotometer, a light source would also be used to illuminate the print sample (#1) but it is mathematically removed from the spectral measurement so the device outputs spectral response data in percentage numbers. This can be used to calculate how the print sample might appear under different lighting conditions (#2 above).
So, with a spectrophotometer you can build profiles for different lighting conditions providing you have measured spectral for those lights.
I hope this helps
-
So, with a spectrophotometer you can build profiles for different lighting conditions providing you have measured spectral for those lights.
This can be done with a spectrophotometer however, I'm not aware of any tool that can generate these kind of specialty profiles directly. I1Profiler does let you generate a profile for a custom light source and a multiple of defined spectra such a "A." But it does not make a print look like it's 5000K illuminated with a 3500K bulb. What it does is correct color deviations from spikey waveforms so that a print will look most natural under that light. An example is in order.
Take the case of cheap, 2700K LEDs v regular 2700K incandescents. The cheap LEDs significantly attenuate reds and cyans even though they may appear to be the same CCT "white." Measuring the spectra of the LEDs then using I1Profiler to make a profile will push the printer to make the cyans and reds more saturated than a normal profile would do. The result is that a print made with the LED spectra profile will look much closer under LED light to that of a regular print under the incandescent. But it will still look warmer than the normal print viewed under 5000K. Even with the (relatively) attenuated reds.
There are ways to achieve what the OP indicates he desires but not directly. For instance one could download the CGATs spectral files from a profile target and go through all the spectral readings and adjust the reflectance values against the ratio of the 3500K/5000K spectra. Then generate a profile from that.
It will accomplish approximately the same result as the curves adjustment I recommended earlier and that requires no spectral info which the Spyder does not provide, nor a more complicated workflow requiring some customization of tools.
-
What I understand from that topic is that for each set of "printer-ink-paper" exists only one color profile which covers а maximum color space (volume). It will give natural colors under light near 5000k or somеwhere in the middle between 1000 and 10000.
It is still possible to be generated specific color profiles for different ColTemp, but they will cover less color space and will not look as a normal print under 5000K.
So, when I want a colder or warmer print, lifted shadows or different blue color, it is better to tweak my picture before printing.
Is that correct?
-
What I understand from that topic is that for each set of "printer-ink-paper" exists only one color profile which covers а maximum color space (volume). It will give natural colors under light near 5000k or somеwhere in the middle between 1000 and 10000.
It is still possible to be generated specific color profiles for different ColTemp, but they will cover less color space and will not look as a normal print under 5000K.
It's possible but only with data from a spectrophotometer following the process I outlined with special tools that you would likely have to make yourself. Since you do not have spectral data it isn't an option.
So, when I want a colder or warmer print, lifted shadows or different blue color, it is better to tweak my picture before printing.
Is that correct?
Yes. I provided a technique to make these adjustments to make an image adjustment so that a print will look, when illuminated with 3500K light, look like it was illuminated with 5000K light.
Just a warning though. If any of the 3500K light illuminates nearby material, unprinted border, or backing other than the print, your eye/brain will adapt and the image will look too blue. There is a complex relationship to actual spectral color (hard numbers) and how "colors" appear and much of it is psychological. You might find tweaking the most productive approach.
-
This can be done with a spectrophotometer however, I'm not aware of any tool that can generate these kind of specialty profiles directly. I1Profiler does let you generate a profile for a custom light source and a multiple of defined spectra such a "A."
AFAIK it (i1P) can, as could the old ProfileMaker pro, no? The instruments used provide spectral data to do so.
-
AFAIK it (i1P) can, as could the old ProfileMaker pro, no? The instruments used provide spectral data to do so.
PM5 and I1P can provide the spectral data but neither provide the tools to rescale the spectral data from 5000K to 3500K. I have tools that easily do it but use them for other purposes. I see few, if any, actual uses for it.
OTOH, accounting for custom illuminant spectral deviations from black body or D50, which the products have, can produce much better looking prints when viewed in the otherwise marginal light. But these are all Bradford adapted from whatever white point the custom illuminant has to D50 and the profile will be tagged as D50.
-
PM5 and I1P can provide the spectral data but neither provide the tools to rescale the spectral data from 5000K to 3500K.
I don't know what you mean by scale. My understanding (and experience with both) is one measured the spectrum of the lighting under which prints will be viewed and that's used instead of the usual assumption of D50. I've seen subtle but visible differences doing this. What and why would rescale be use differently?
-
I don't know what you mean by scale. My understanding (and experience with both) is one measured the spectrum of the lighting under which prints will be viewed and that's used instead of the usual assumption of D50. I've seen subtle but visible differences doing this. What and why would rescale be use differently?
Rescaling is needed so that the OP can view a print under 3500K and have it look like a normal print illuminated with 5000K side by side. In other words the print, when viewed under 3500K should look like a soft proof on a monitor profiled to 5000K side by side.
A rather unusual thing to do but I can see making something like that for creating special effects.
The stuff built in to I1P does a good job of making a print that, when viewed under a crummy 2700K LED will look like a normal print viewed under 2700K tungsten.
These are vastly different things.
Added:
Rescaling would be done by taking the reflectance at each wavelength and multiplying by the ratio of 5000K and 3500K at the same wavelength. The result would need to be scaled by a constant such that the max reflectance is no more than 1.0
-
Rescaling is needed so that the OP can view a print under 3500K and have it look like a normal print illuminated with 5000K side by side. In other words the print, when viewed under 3500K should look like a soft proof on a monitor profiled to 5000K side by side.
That seems to be exactly what the features in i1P and PMP were supposed to do I thought. Again I don't understand this bit about scaling. But my understanding and experience is, you have a light source of which you're to be viewing your prints. You measure it and use that with the profile instead of D50.
My understanding and this person sync up:
http://nativedigital.blogspot.com/2011/03/rgb-printer-profiling-with-i1-profiler.html (http://nativedigital.blogspot.com/2011/03/rgb-printer-profiling-with-i1-profiler.html)
The next page, Lighting, is a new option that will be unfamiliar to users of i1 Match. Printer profiles can be built based on an expected viewing light to help improve the accuracy of the profile under that lighting condition. For example, if you were producing prints for a gallery that had a particular type of lights you could measure those lights and then use that measurement data when the profile was created to make the prints look good under that light source. Usually, however, you don’t know where the print will be viewed so the best thing for the vast majority of users is to leave the lighting options set to the default of D50 and skip the Lighting page.
What the OP wants to do is moot, his measuring device isn't a Spectrophotometer.
-
https://www.youtube.com/watch?v=ijDsdX-EtZ4
Go to minute 21.
-
That seems to be exactly what the features in i1P and PMP were supposed to do I thought. Again I don't understand this bit about scaling. But my understanding and experience is, you have a light source of which you're to be viewing your prints. You measure it and use that with the profile instead of D50.
My understanding and this person sync up:
http://nativedigital.blogspot.com/2011/03/rgb-printer-profiling-with-i1-profiler.html (http://nativedigital.blogspot.com/2011/03/rgb-printer-profiling-with-i1-profiler.html)
The next page, Lighting, is a new option that will be unfamiliar to users of i1 Match. Printer profiles can be built based on an expected viewing light to help improve the accuracy of the profile under that lighting condition. For example, if you were producing prints for a gallery that had a particular type of lights you could measure those lights and then use that measurement data when the profile was created to make the prints look good under that light source. Usually, however, you don’t know where the print will be viewed so the best thing for the vast majority of users is to leave the lighting options set to the default of D50 and skip the Lighting page.
All that is correct but not what the OP apparently desires. His initial question was somewhat ambiguous so I asked some further questions. He specifically indicated that he wanted a print that when illuminated with 3500K would look close to the same image on a monitor, presumably at 5000K he referenced. And he is looking at both at the same time. See my answer #6 in the thread.
The custom illuminant option only improves the color as it appears after adaptation. That is it improves color after one is adaptated to the illuminant's white. It does not make the print appear the same as the monitor at 5000K when looked at side by side. It doesn't change the white point or other neutral colors materially. Even using Abs. Col. They remain just as reddish viewed under 3500K as neutrals from a regular print profile.
What the OP wants to do is moot, his measuring device isn't a Spectrophotometer.
Right. That's why I suggested he achieve approximately the same results by lowering the RG curve end points as described in reply #8. Not as good but should be reasonable and is his only option w/o a spectro.
It is, however, quite difficult to illuminant such a print with 3500K without leaking the light to other things ruining adaptation.
-
https://www.youtube.com/watch?v=ijDsdX-EtZ4
Go to minute 21.
I have no qualms at all with the video but I don't think it addresses what the OP wants to achieve. This is all quite confused due to adaptation so let me propose a scenario that gets that out of the equation.
Take an image that is reasonably in gamut on both screen and printer. Create a print using a normal D50 profile. Call that PrintD50. Now make a print using a profile made with a custom illuminant that's 3500K. Call that Print35C. Make a third print using either the scaled spectral approach outlined or by shifting the RG downward in curves per reply #8. Call that Print35S
Now illuminate PrintD50, Print35C, and Print35S with 3500K.
Take a RAW picture of each of these prints and the monitor which is profiled to D50. Process all adjusting for the same luminance but maintaining identical temperature and g/m offset which are set to match the image of the monitor. Paste them all on a single page. Adaptation and even cognitive effects are no longer a factor. L/I issues with the camera imager is a factor, but small in comparison to the relatively stark differences in color temp.
The monitor will be reasonably close to Print35S but Print35C and PrintD50 will both be quite reddish and much closer to each other than either the monitor or Print35S.
I hope this provides more clarity.
Edited to add:
My perspective on what I1P accomplishes with custom illuminant profiles is that they produce optimal profiles for the illuminant when a person is adaptated to that illuminant. This is, by far, the most common use. When such a print is displayed under that illuminant, such as in an office under a 2700K, 3500K or 4000K light, the person viewing will normally be adaptated to the office illuminant and the print made with a custom illuminant will look proper and produce better color than one with a standard D50 profile. But it's not a huge effect.
A print made to reflect light such that it appears as if illuminated with D50 in a lower CCT environment will look quite strange and very bluish because the viewer will not be adaptated to D50. This is why I stated that doing something like the OP appears to want is appropriate only for some specialized purpose like creating an illusion or special effect. It's not a good idea for general printing.
-
Thanks for your input to this discussion. It has been very useful.