Colorchecker calibration, Gamma and Other correction techniques

Sdiver2489 · Jul 29, 2022

Hello all,

I've been getting more and more into photographic reproduction of artwork. What started as a primarily composite reconstruction of Xerox scane and trial and error color correction has now become a dedicated photographic setup with external lights, soon to be polarized light sources and filters, pixel shift images and more.

With each painting I do, I try to get that much closer to perfection. What use to take me many iterations now usually only takes a couple to get near perfect.

My latest experimentation comes from not only using a colorchecker passport but also using the swatches on the passport to calibrate gamma.

Now, this is where the theory is bit perhaps beyond me at the moment but looking to learn from you all. I take a picture of the colorchecker and the x-rite software generates a DNG profile for the image. When I apply that image I primarily see changes in hue and saturation of certain colors. After this I then white balance the image using the white balance target on the colorchecker.

What eluded me for a while is how to standardize exposure and contrast/gamma curves. What I then did was set my exposure to set 18% gray at roughly 50% luminosity. I then went and in the tone curve section of lightroom, set it to a point curve and selected each grayscale swatch and moved it to the luminosity value for that swatch provided by X-rite. To take it a step further, I then went and went into each individual RGB tone curve and equalized the tone curves for each grayscale swatch.

The results are very solid but I want to make sure I'm not making a mistake that just happens to work out nicely. The main questions that come to mind are:

1. If the colorchecker has these nice grayscale swatches, why doesn't it correct for contrast as well it seems? If the goal is to arrive at reality why am I going in manually and adjusting the tone curve to match the standard swatches?

2. If I then go and correct the RGB tone curves am I making the grayscale swatches more accurate at the expense of colors?

Any other tips for how to arrive at the perfect color/contrast reproduction of artwork is appreciated. Thanks all!

Erik Kaffehr · Jul 30, 2022

Sdiver2489 said:
Hello all,

I've been getting more and more into photographic reproduction of artwork. What started as a primarily composite reconstruction of Xerox scane and trial and error color correction has now become a dedicated photographic setup with external lights, soon to be polarized light sources and filters, pixel shift images and more.

With each painting I do, I try to get that much closer to perfection. What use to take me many iterations now usually only takes a couple to get near perfect.

My latest experimentation comes from not only using a colorchecker passport but also using the swatches on the passport to calibrate gamma.

Now, this is where the theory is bit perhaps beyond me at the moment but looking to learn from you all. I take a picture of the colorchecker and the x-rite software generates a DNG profile for the image. When I apply that image I primarily see changes in hue and saturation of certain colors. After this I then white balance the image using the white balance target on the colorchecker.

What eluded me for a while is how to standardize exposure and contrast/gamma curves. What I then did was set my exposure to set 18% gray at roughly 50% luminosity. I then went and in the tone curve section of lightroom, set it to a point curve and selected each grayscale swatch and moved it to the luminosity value for that swatch provided by X-rite. To take it a step further, I then went and went into each individual RGB tone curve and equalized the tone curves for each grayscale swatch.

The results are very solid but I want to make sure I'm not making a mistake that just happens to work out nicely. The main questions that come to mind are:

1. If the colorchecker has these nice grayscale swatches, why doesn't it correct for contrast as well it seems? If the goal is to arrive at reality why am I going in manually and adjusting the tone curve to match the standard swatches?

2. If I then go and correct the RGB tone curves am I making the grayscale swatches more accurate at the expense of colors?

Any other tips for how to arrive at the perfect color/contrast reproduction of artwork is appreciated. Thanks all!

Hi,

The tone curve is needed to match luminance range to media. Real subjects can have considerable luminance range but prints often have quite limited density range.

So, tone curves are used to map different luminance ranges onto media with a more limited range.

With repro, you would probably start out with a subject with a very limited luminance range that you would reproduce on medium with quite limited luminance (or density) range.

To do that, you would need a 'repro profile' that is essentially linear.

The tool I use to generate profiles is 'Lumariver Profile Designer': http://www.lumariver.com/lrpd-manual/

Best regards

Erik

ggbutcher · Jul 30, 2022

Erik Kaffehr said:
Sdiver2489 said:

Hello all,

I've been getting more and more into photographic reproduction of artwork. What started as a primarily composite reconstruction of Xerox scane and trial and error color correction has now become a dedicated photographic setup with external lights, soon to be polarized light sources and filters, pixel shift images and more.

With each painting I do, I try to get that much closer to perfection. What use to take me many iterations now usually only takes a couple to get near perfect.

My latest experimentation comes from not only using a colorchecker passport but also using the swatches on the passport to calibrate gamma.

Now, this is where the theory is bit perhaps beyond me at the moment but looking to learn from you all. I take a picture of the colorchecker and the x-rite software generates a DNG profile for the image. When I apply that image I primarily see changes in hue and saturation of certain colors. After this I then white balance the image using the white balance target on the colorchecker.

What eluded me for a while is how to standardize exposure and contrast/gamma curves. What I then did was set my exposure to set 18% gray at roughly 50% luminosity. I then went and in the tone curve section of lightroom, set it to a point curve and selected each grayscale swatch and moved it to the luminosity value for that swatch provided by X-rite. To take it a step further, I then went and went into each individual RGB tone curve and equalized the tone curves for each grayscale swatch.

The results are very solid but I want to make sure I'm not making a mistake that just happens to work out nicely. The main questions that come to mind are:

1. If the colorchecker has these nice grayscale swatches, why doesn't it correct for contrast as well it seems? If the goal is to arrive at reality why am I going in manually and adjusting the tone curve to match the standard swatches?

2. If I then go and correct the RGB tone curves am I making the grayscale swatches more accurate at the expense of colors?

Any other tips for how to arrive at the perfect color/contrast reproduction of artwork is appreciated. Thanks all!

Click to expand...

Hi,

The tone curve is needed to match luminance range to media. Real subjects can have considerable luminance range but prints often have quite limited density range.

Yes. The best way IMHO to think of tone mapping is that the camera captured tone values that represent the energy in the scene, and the tone mapping required for viewing will depend on the particular media.

Color management should make that consistent across media. Calibrating/profiling a display includes a tone curve designed to take the data from "linear" to the response of the display; same thing for a printer profile. So, the tone curve you apply in the raw processor or photo editor prior to any rendition profile will then operate consistently across all the media.

Erik Kaffehr said:
So, tone curves are used to map different luminance ranges onto media with a more limited range.

With repro, you would probably start out with a subject with a very limited luminance range that you would reproduce on medium with quite limited luminance (or density) range.

To do that, you would need a 'repro profile' that is essentially linear.

Erik Kaffehr said:
The tool I use to generate profiles is 'Lumariver Profile Designer

+1

Erik Kaffehr said:
Best regards

Erik

Jack Hogan · Jul 31, 2022

Sdiver2489 said:
Sdiver2489 wrote: [...] My latest experimentation comes from not only using a colorchecker passport but also using the swatches on the passport to calibrate gamma. [...]

Hello Sdiver2489, welcome to the forum.

as Erik and Glenn have mentioned, color transformation is ideally supposed to be wholly linear. Historically, though, displays had a native power law response which needed to be counteracted by dialing in a corresponding pre-emphasis, referred in these pages as 'gamma', to cancel it out. So the sensor to eye trip is/was ideally linear:

light --> raw -----> XYZ -> RGB -> RGB^0.45 -> RGB^2.2 -----> light to eyes
lin -----> lin -------> lin ---> lin ----> 'gamma' --> 'de-gamma' -> lin
ether -> sensor-> color trans --> ...sRGB... --> monitor ------> eyes

With a well calibrated monitor the gamma and de-gamma steps should cancel each other out: (RGB^0.4545)^2.2 = RGB.

A colorchecker target with proper software (+1 for Lumariver) helps make sure that the linear trip from raw to XYZ is as good as it can be for your illuminant, camera and lens. Having achieved that, any 'gamma' changes you make with an editor (say by playing with curves as I believe you suggest above) messes with the ideal gamma/de-gamma function, giving you an arbitrary result. Of course if the messing is minimal, so will be the end effect.

With repro work your time would be better spent finding a Tone Mapping Operator (TMO or TRO) appropriate for your expected viewing device and conditions. As Erik and Glenn say, the objective is to squeeze the wider dynamic range of the raw capture into the smaller one of the output device as set up - with minimal perceptual damage.

Jack

Sdiver2489 · Jul 31, 2022

Jack Hogan said:
Jack Hogan said:

Sdiver2489 wrote: [...] My latest experimentation comes from not only using a colorchecker passport but also using the swatches on the passport to calibrate gamma. [...]

Click to expand...

Hello Sdiver2489, welcome to the forum.

as Erik and Glenn have mentioned, color transformation is ideally supposed to be wholly linear. Historically, though, displays had a native power law response which needed to be counteracted by dialing in a corresponding pre-emphasis, referred in these pages as 'gamma', to cancel it out. So the sensor to eye trip is/was ideally linear:

light --> raw -----> XYZ -> RGB -> RGB^0.45 -> RGB^2.2 -----> light to eyes
lin -----> lin -------> lin ---> lin ----> 'gamma' --> 'de-gamma' -> lin
ether -> sensor-> color trans --> ...sRGB... --> monitor ------> eyes

With a well calibrated monitor the gamma and de-gamma steps should cancel each other out: (RGB^0.4545)^2.2 = RGB.

A colorchecker target with proper software (+1 for Lumariver) helps make sure that the linear trip from raw to XYZ is as good as it can be for your illuminant, camera and lens. Having achieved that, any 'gamma' changes you make with an editor (say by playing with curves as I believe you suggest above) messes with the ideal gamma/de-gamma function, giving you an arbitrary result. Of course if the messing is minimal, so will be the end effect.

With repro work your time would be better spent finding a Tone Mapping Operator (TMO or TRO) appropriate for your expected viewing device and conditions. As Erik and Glenn say, the objective is to squeeze the wider dynamic range of the raw capture into the smaller one of the output device as set up - with minimal perceptual damage.

Jack

Hi Jack,

This is all interesting information and honestly a little above my thought process right now so I definitely need to delve deeper into this.

Can you help correct where my thought process might be wrong? Here is my basic thought process:

1. The painting has a certain dynamic range/contrast and exposure that I want to replicate

2. The camera RAW data does not necessary match that same dynamic range

3. The X-rite profiler corrects for hue and saturation differences in the captured image versus reality

4. The X-rite profiler makes no attempt to correct the exposure or the tone curve/contrast of the image to match that of the subject.

5. I know some information via the swatches in terms of their L*a*b color values that should be replicated in the captured image. Namely, 18% grey should be nominally at 50% of my histogram

6. There are published lab values for each of the other swatches on the colorchecker. If I manually correct the tone curve for these values...aren't I basically making the gamma correction necessary to match the print based on the subject?

It seems odd that, via this method, my prints have been pretty well nailed on the first attempt if it truly was just arbitrary.

I really don't mind being wrong I'm just trying to see what magic Lumariver is doing that can't be done "by hand" when its using the same data via the analysis of the colorchecker image.

The only reasons I'm resistant to getting lumariver is:

a. It appears to be $200 for a reproduction profile license and I don't make money on this. This is just a hobby.

b. It doesn't let me trial run a profile. If I could, and I saw the results were noticeably better than what I got via the method above...I might determine its necessary.

Any feedback you all have is appreciated and my apologies for my basic understanding. The method I describe above was just the solution I came up with in order to solve the issue with "calibrating" my captured image to print to the same values as the original rather than just guess and check.

Erik Kaffehr · Aug 1, 2022

Sdiver2489 said:
Jack Hogan said:

Sdiver2489 said:

Sdiver2489 wrote: [...] My latest experimentation comes from not only using a colorchecker passport but also using the swatches on the passport to calibrate gamma. [...]

Click to expand...

Hello Sdiver2489, welcome to the forum.

as Erik and Glenn have mentioned, color transformation is ideally supposed to be wholly linear. Historically, though, displays had a native power law response which needed to be counteracted by dialing in a corresponding pre-emphasis, referred in these pages as 'gamma', to cancel it out. So the sensor to eye trip is/was ideally linear:

light --> raw -----> XYZ -> RGB -> RGB^0.45 -> RGB^2.2 -----> light to eyes
lin -----> lin -------> lin ---> lin ----> 'gamma' --> 'de-gamma' -> lin
ether -> sensor-> color trans --> ...sRGB... --> monitor ------> eyes

With a well calibrated monitor the gamma and de-gamma steps should cancel each other out: (RGB^0.4545)^2.2 = RGB.

A colorchecker target with proper software (+1 for Lumariver) helps make sure that the linear trip from raw to XYZ is as good as it can be for your illuminant, camera and lens. Having achieved that, any 'gamma' changes you make with an editor (say by playing with curves as I believe you suggest above) messes with the ideal gamma/de-gamma function, giving you an arbitrary result. Of course if the messing is minimal, so will be the end effect.

With repro work your time would be better spent finding a Tone Mapping Operator (TMO or TRO) appropriate for your expected viewing device and conditions. As Erik and Glenn say, the objective is to squeeze the wider dynamic range of the raw capture into the smaller one of the output device as set up - with minimal perceptual damage.

Jack

Click to expand...

Hi Jack,

This is all interesting information and honestly a little above my thought process right now so I definitely need to delve deeper into this.

Can you help correct where my thought process might be wrong? Here is my basic thought process:

1. The painting has a certain dynamic range/contrast and exposure that I want to replicate

2. The camera RAW data does not necessary match that same dynamic range

It will be a nearly exact match, as long as the luminance range of the subject is within the DR of the sensor. Doing repro, that will be mostly the case.

Sdiver2489 said:
3. The X-rite profiler corrects for hue and saturation differences in the captured image versus reality

No, it just produces a set of color matrices, adds some saturation and adds a tone curve. The way it applies the tone curve may cause color shifts.

Sdiver2489 said:
4. The X-rite profiler makes no attempt to correct the exposure or the tone curve/contrast of the image to match that of the subject.

No, that cannot be done at the profile level. The way to that in Lightroom is to use 'soft profing'. Soft proofing can 'simulate' effects of both ink and paper surface.

Sdiver2489 said:
5. I know some information via the swatches in terms of their L*a*b color values that should be replicated in the captured image. Namely, 18% grey should be nominally at 50% of my histogram

May be, with a linear profile it may not matter.

Sdiver2489 said:
6. There are published lab values for each of the other swatches on the colorchecker. If I manually correct the tone curve for these values...aren't I basically making the gamma correction necessary to match the print based on the subject?

No, you would just add bends to the tone curve.

Sdiver2489 said:
It seems odd that, via this method, my prints have been pretty well nailed on the first attempt if it truly was just arbitrary.

Could be...

Sdiver2489 said:
I really don't mind being wrong I'm just trying to see what magic Lumariver is doing that can't be done "by hand" when its using the same data via the analysis of the colorchecker image.

LumaRiver does not do any magic, but it has an option to create a 'reproduction' profile, that applies minimal corrections to the tone curve.

Sdiver2489 said:
The only reasons I'm resistant to getting lumariver is:

a. It appears to be $200 for a reproduction profile license and I don't make money on this. This is just a hobby.

There are different versions at different cost, but the one having the option to create a repro profile.

Sdiver2489 said:
b. It doesn't let me trial run a profile. If I could, and I saw the results were noticeably better than what I got via the method above...I might determine its necessary.

Any feedback you all have is appreciated and my apologies for my basic understanding. The method I describe above was just the solution I came up with in order to solve the issue with "calibrating" my captured image to print to the same values as the original rather than just guess and check.

A simple way to fake a linear profile in Lightroom is to switch to 'Process Version 2' in 'Develop->Settings' and switch to 'tone curve->linear. Switching back to 'Process Version 5' adds a tone curve that inverts the default tone curve.

This compares Adobe Color with Lumariver generated profile. The image is split across one diagonal.

This shows the Lumariver Reproduction Profile vs the 'linearized' Lumariver Profile.

I would recall that the Adobe standard and Color profiles are pretty accurate. So, linearizing the Adobe Color profile may be a good option.

Soft proofing in Lightroom shows how the image will render in print. In this case the image will fit well in the color space of the print. I would use relative rendering indent.

Best regards

Erik

--
Erik Kaffehr
Website: http://echophoto.dnsalias.net
Magic uses to disappear in controlled experiments…
Gallery: http://echophoto.smugmug.com
Articles: http://echophoto.dnsalias.net/ekr/index.php/photoarticles

ggbutcher · Aug 1, 2022

Sdiver2489 said:
1. The painting has a certain dynamic range/contrast and exposure that I want to replicate

2. The camera RAW data does not necessary match that same dynamic range

Gonna lay this out maybe more for my own clarification...

The dynamic range of the painting itself is really more about the reflectance of the pigments than anything, so an even lighting should provide a pretty low dynamic range. I mean, the painter decided how to render the dynamic range of the scene, you're stuck with that in the pigments, no?

So then, the tone curve is all about accommodating any perceptual differences at rendition and not at all about compressing dynamic range. Heck, I haven't done any serious repro, but I'm wondering if any departure-from-linear tone curve is really needed, except for the one that is in the display/export profile... ??

Jack Hogan · Aug 1, 2022

Sdiver2489 said:
1. The painting has a certain dynamic range/contrast and exposure that I want to replicate

The painting reflects the illuminant ideally diffusely towards the camera, producing an image on the sensing plane, captured in the raw data by the sensor. This is a linear process.

Sdiver2489 said:
2. The camera RAW data does not necessary match that same dynamic range

Assuming diffuse reflection, proper exposure and a typical painting, image intensity is captured appropriately in the raw data

Sdiver2489 said:
3. The X-rite profiler corrects for hue and saturation differences in the captured image versus reality

The profile should ensure that image data is projected to the Profile Connection Space (XYZ) as accurately as possible, so that output colors will be as accurate as possible

Sdiver2489 said:
4. The X-rite profiler makes no attempt to correct the exposure or the tone curve/contrast of the image to match that of the subject.

It shouldn't, but most profiles add a generic tone curve to compensate for a more limited DR in the display device and produce a more pleasing output image. This is almost always incorrect and, as Glenn says, you may not need it at all in your application. Lumariver/DcamProf provide the most options, hence the best solutions, in this critical step.

Sdiver2489 said:
5. I know some information via the swatches in terms of their L*a*b color values that should be replicated in the captured image. Namely, 18% grey should be nominally at 50% of my histogram

It depends on how bright the brightest highlights you would like to preserve detail in are

Sdiver2489 said:
6. There are published lab values for each of the other swatches on the colorchecker. If I manually correct the tone curve for these values...aren't I basically making the gamma correction necessary to match the print based on the subject?

Forget about gamma, that's not what needs to be corrected assuming all else is working as it should in the profile. The job of a good profile is to make sure the colors are accurate. When you mess with curves via an editor, you introduce unwanted changes in chromaticity. Such changes may not be visible in the neutral tones and be pleasing to the eye - but the end result is less accurate.

Sdiver2489 said:
It seems odd that, via this method, my prints have been pretty well nailed on the first attempt if it truly was just arbitrary.

I really don't mind being wrong I'm just trying to see what magic Lumariver is doing that can't be done "by hand" when its using the same data via the analysis of the colorchecker image.

The only reasons I'm resistant to getting lumariver is:

a. It appears to be $200 for a reproduction profile license and I don't make money on this. This is just a hobby.

b. It doesn't let me trial run a profile. If I could, and I saw the results were noticeably better than what I got via the method above...I might determine its necessary.

Any feedback you all have is appreciated and my apologies for my basic understanding. The method I describe above was just the solution I came up with in order to solve the issue with "calibrating" my captured image to print to the same values as the original rather than just guess and check.

The objective is to minimize the errors in deltaE in each of the 24 colorchecker patches. What are they with your method? If the average of the 18 color patches is less than about 1 dE00 and you have no critical patches greater than about 2 you are doing well.

Even better make up a target with, or measure the spectral reflectance of, the pigments of your paintings and use those to produce a profile to measure against.

Jack

Jack Hogan · Aug 1, 2022

ggbutcher said:
Sdiver2489 said:

1. The painting has a certain dynamic range/contrast and exposure that I want to replicate

2. The camera RAW data does not necessary match that same dynamic range

Click to expand...

Gonna lay this out maybe more for my own clarification...

The dynamic range of the painting itself is really more about the reflectance of the pigments than anything, so an even lighting should provide a pretty low dynamic range. I mean, the painter decided how to render the dynamic range of the scene, you're stuck with that in the pigments, no?

So then, the tone curve is all about accommodating any perceptual differences at rendition and not at all about compressing dynamic range. Heck, I haven't done any serious repro, but I'm wondering if any departure-from-linear tone curve is really needed, except for the one that is in the display/export profile... ??

Good point Glenn

D Cox · Aug 2, 2022

Jack Hogan said:
Sdiver2489 said:

1. The painting has a certain dynamic range/contrast and exposure that I want to replicate

Click to expand...

The painting reflects the illuminant ideally diffusely towards the camera, producing an image on the sensing plane, captured in the raw data by the sensor. This is a linear process.

Jack Hogan said:

2. The camera RAW data does not necessary match that same dynamic range

Click to expand...

Assuming diffuse reflection, proper exposure and a typical painting, image intensity is captured appropriately in the raw data

Jack Hogan said:

3. The X-rite profiler corrects for hue and saturation differences in the captured image versus reality

Click to expand...

The profile should ensure that image data is projected to the Profile Connection Space (XYZ) as accurately as possible, so that output colors will be as accurate as possible

Jack Hogan said:

4. The X-rite profiler makes no attempt to correct the exposure or the tone curve/contrast of the image to match that of the subject.

Click to expand...

It shouldn't, but most profiles add a generic tone curve to compensate for a more limited DR in the display device and produce a more pleasing output image. This is almost always incorrect and, as Glenn says, you may not need it at all in your application. Lumariver/DcamProf provide the most options, hence the best solutions, in this critical step.

Jack Hogan said:

5. I know some information via the swatches in terms of their L*a*b color values that should be replicated in the captured image. Namely, 18% grey should be nominally at 50% of my histogram

Click to expand...

It depends on how bright the brightest highlights you would like to preserve detail in are

Jack Hogan said:

6. There are published lab values for each of the other swatches on the colorchecker. If I manually correct the tone curve for these values...aren't I basically making the gamma correction necessary to match the print based on the subject?

Click to expand...

Forget about gamma, that's not what needs to be corrected assuming all else is working as it should in the profile. The job of a good profile is to make sure the colors are accurate. When you mess with curves via an editor, you introduce unwanted changes in chromaticity. Such changes may not be visible in the neutral tones and be pleasing to the eye - but the end result is less accurate.

Jack Hogan said:

It seems odd that, via this method, my prints have been pretty well nailed on the first attempt if it truly was just arbitrary.

I really don't mind being wrong I'm just trying to see what magic Lumariver is doing that can't be done "by hand" when its using the same data via the analysis of the colorchecker image.

The only reasons I'm resistant to getting lumariver is:

a. It appears to be $200 for a reproduction profile license and I don't make money on this. This is just a hobby.

b. It doesn't let me trial run a profile. If I could, and I saw the results were noticeably better than what I got via the method above...I might determine its necessary.

Any feedback you all have is appreciated and my apologies for my basic understanding. The method I describe above was just the solution I came up with in order to solve the issue with "calibrating" my captured image to print to the same values as the original rather than just guess and check.

Click to expand...

The objective is to minimize the errors in deltaE in each of the 24 colorchecker patches. What are they with your method? If the average of the 18 color patches is less than about 1 dE00 and you have no critical patches greater than about 2 you are doing well.

Even better make up a target with, or measure the spectral reflectance of, the pigments of your paintings and use those to produce a profile to measure against.

Jack

In practice, a .dcp profile made from the Color Checker gives a good rendering of artists' pigments.

Don

Entropy512 · Aug 2, 2022

Erik Kaffehr said:
No, it just produces a set of color matrices, adds some saturation and adds a tone curve. The way it applies the tone curve may cause color shifts.

No tone curve for X-Rite's software, at least as of 2-3 years ago. Just a basic dual-illuminant matrix. No LUTs (not even the "2.5D" LUTs used by Adobe and dcamprof), no tone curve.

Entropy512 · Aug 2, 2022

ggbutcher said:
Sdiver2489 said:

1. The painting has a certain dynamic range/contrast and exposure that I want to replicate

2. The camera RAW data does not necessary match that same dynamic range

Click to expand...

Gonna lay this out maybe more for my own clarification...

The dynamic range of the painting itself is really more about the reflectance of the pigments than anything, so an even lighting should provide a pretty low dynamic range. I mean, the painter decided how to render the dynamic range of the scene, you're stuck with that in the pigments, no?

So then, the tone curve is all about accommodating any perceptual differences at rendition and not at all about compressing dynamic range. Heck, I haven't done any serious repro, but I'm wondering if any departure-from-linear tone curve is really needed, except for the one that is in the display/export profile... ??

Gamut is more likely to be an issue here. Dynamic range is unlikely to be a problem for an evenly lit painting, but some pigments likely land outside of the extremely limited sRGB gamut.

Jack Hogan · Aug 2, 2022

Entropy512 said:
Erik Kaffehr said:

No, it just produces a set of color matrices, adds some saturation and adds a tone curve. The way it applies the tone curve may cause color shifts.

Click to expand...

No tone curve for X-Rite's software, at least as of 2-3 years ago. Just a basic dual-illuminant matrix. No LUTs (not even the "2.5D" LUTs used by Adobe and dcamprof), no tone curve.

Interesting, I never used it but without a tone curve related output would typically look flat. Does it produce a dcp or ICC input profile that can be looked at?

D Cox · Aug 3, 2022

Entropy512 said:
ggbutcher said:

Sdiver2489 said:

1. The painting has a certain dynamic range/contrast and exposure that I want to replicate

2. The camera RAW data does not necessary match that same dynamic range

Click to expand...

Gonna lay this out maybe more for my own clarification...

The dynamic range of the painting itself is really more about the reflectance of the pigments than anything, so an even lighting should provide a pretty low dynamic range. I mean, the painter decided how to render the dynamic range of the scene, you're stuck with that in the pigments, no?

So then, the tone curve is all about accommodating any perceptual differences at rendition and not at all about compressing dynamic range. Heck, I haven't done any serious repro, but I'm wondering if any departure-from-linear tone curve is really needed, except for the one that is in the display/export profile... ??

Click to expand...

Gamut is more likely to be an issue here. Dynamic range is unlikely to be a problem for an evenly lit painting, but some pigments likely land outside of the extremely limited sRGB gamut.

The people who make pigments and dyes do try to include in their range of products the most saturated colours they can make.

Don

Erik Kaffehr · Aug 3, 2022

D Cox said:
Entropy512 said:

ggbutcher said:

Sdiver2489 said:

1. The painting has a certain dynamic range/contrast and exposure that I want to replicate

2. The camera RAW data does not necessary match that same dynamic range

Click to expand...

Gonna lay this out maybe more for my own clarification...

The dynamic range of the painting itself is really more about the reflectance of the pigments than anything, so an even lighting should provide a pretty low dynamic range. I mean, the painter decided how to render the dynamic range of the scene, you're stuck with that in the pigments, no?

So then, the tone curve is all about accommodating any perceptual differences at rendition and not at all about compressing dynamic range. Heck, I haven't done any serious repro, but I'm wondering if any departure-from-linear tone curve is really needed, except for the one that is in the display/export profile... ??

Click to expand...

Gamut is more likely to be an issue here. Dynamic range is unlikely to be a problem for an evenly lit painting, but some pigments likely land outside of the extremely limited sRGB gamut.

Click to expand...

The people who make pigments and dyes do try to include in their range of products the most saturated colours they can make.

Don

I don't know about that...

But, I don't think it is the task of the profile to map the 'RGB' signals into a small color space.

Best regards

Erik

D Cox · Aug 3, 2022

Erik Kaffehr said:
D Cox said:

Entropy512 said:

ggbutcher said:

Sdiver2489 said:

1. The painting has a certain dynamic range/contrast and exposure that I want to replicate

2. The camera RAW data does not necessary match that same dynamic range

Click to expand...

Gonna lay this out maybe more for my own clarification...

The dynamic range of the painting itself is really more about the reflectance of the pigments than anything, so an even lighting should provide a pretty low dynamic range. I mean, the painter decided how to render the dynamic range of the scene, you're stuck with that in the pigments, no?

So then, the tone curve is all about accommodating any perceptual differences at rendition and not at all about compressing dynamic range. Heck, I haven't done any serious repro, but I'm wondering if any departure-from-linear tone curve is really needed, except for the one that is in the display/export profile... ??

Click to expand...

Gamut is more likely to be an issue here. Dynamic range is unlikely to be a problem for an evenly lit painting, but some pigments likely land outside of the extremely limited sRGB gamut.

Click to expand...

The people who make pigments and dyes do try to include in their range of products the most saturated colours they can make.

Don

Click to expand...

I don't know about that...

But, I don't think it is the task of the profile to map the 'RGB' signals into a small color space.

Best regards

Erik

I agree.

Don

Entropy512 · Aug 3, 2022

Jack Hogan said:
Entropy512 said:

Erik Kaffehr said:

No, it just produces a set of color matrices, adds some saturation and adds a tone curve. The way it applies the tone curve may cause color shifts.

Click to expand...

No tone curve for X-Rite's software, at least as of 2-3 years ago. Just a basic dual-illuminant matrix. No LUTs (not even the "2.5D" LUTs used by Adobe and dcamprof), no tone curve.

Click to expand...

Interesting, I never used it but without a tone curve related output would typically look flat. Does it produce a dcp or ICC input profile that can be looked at?

DCP - I haven't used it in ages because dcamprof was so superior.

I don't know how Adobe behaves when the DCP profile lacks a tone curve - maybe when it is completely missing (as opposed to present but indicating linear mapping) it defaults to something not linear?

RawTherapee will default to ignoring the built in tone curve (although currently defaults to using a HueSatMap entry that is tied to tone curve, I've filed an issue to work on making this more consistent after 5.9 gets out - https://github.com/Beep6581/RawTherapee/issues/6467 )

Entropy512 · Aug 3, 2022

Erik Kaffehr said:
D Cox said:

Entropy512 said:

ggbutcher said:

Sdiver2489 said:

1. The painting has a certain dynamic range/contrast and exposure that I want to replicate

2. The camera RAW data does not necessary match that same dynamic range

Click to expand...

Gonna lay this out maybe more for my own clarification...

The dynamic range of the painting itself is really more about the reflectance of the pigments than anything, so an even lighting should provide a pretty low dynamic range. I mean, the painter decided how to render the dynamic range of the scene, you're stuck with that in the pigments, no?

So then, the tone curve is all about accommodating any perceptual differences at rendition and not at all about compressing dynamic range. Heck, I haven't done any serious repro, but I'm wondering if any departure-from-linear tone curve is really needed, except for the one that is in the display/export profile... ??

Click to expand...

Gamut is more likely to be an issue here. Dynamic range is unlikely to be a problem for an evenly lit painting, but some pigments likely land outside of the extremely limited sRGB gamut.

Click to expand...

The people who make pigments and dyes do try to include in their range of products the most saturated colours they can make.

Don

Click to expand...

I don't know about that...

But, I don't think it is the task of the profile to map the 'RGB' signals into a small color space.

Best regards

Erik

True, but the discussion did start moving beyond just input profiles and to tonemapping too (most people think of tonemapping primarily for luminance, but gamut mapping also falls into the generic category of "tonemapping"). It's my opinion (and I think this is shared) that the input profile should not do tonemapping and gamut mapping, leaving that to be done later (as necessary) in the processing pipeline.

Sdiver2489 · Aug 3, 2022

Entropy512 said:
Jack Hogan said:

Entropy512 said:

Erik Kaffehr said:

No, it just produces a set of color matrices, adds some saturation and adds a tone curve. The way it applies the tone curve may cause color shifts.

Click to expand...

No tone curve for X-Rite's software, at least as of 2-3 years ago. Just a basic dual-illuminant matrix. No LUTs (not even the "2.5D" LUTs used by Adobe and dcamprof), no tone curve.

Click to expand...

Interesting, I never used it but without a tone curve related output would typically look flat. Does it produce a dcp or ICC input profile that can be looked at?

Click to expand...

DCP - I haven't used it in ages because dcamprof was so superior.

I don't know how Adobe behaves when the DCP profile lacks a tone curve - maybe when it is completely missing (as opposed to present but indicating linear mapping) it defaults to something not linear?

RawTherapee will default to ignoring the built in tone curve (although currently defaults to using a HueSatMap entry that is tied to tone curve, I've filed an issue to work on making this more consistent after 5.9 gets out - https://github.com/Beep6581/RawTherapee/issues/6467 )

Interesting! I have never heard of dcamprof until now. I actually had the thought that surely there must be some open source software more advanced than x-rites own. Can I hear more about your experience with the software? I have been frustrated that even with a 61MP capture of the Passport, their software will either fail or generate an awful profile if the passport doesn't take up the majority of the frame. Curious if this other software can do better.

ggbutcher · Aug 3, 2022

Sdiver2489 said:
Entropy512 said:

Jack Hogan said:

Entropy512 said:

Erik Kaffehr said:

No, it just produces a set of color matrices, adds some saturation and adds a tone curve. The way it applies the tone curve may cause color shifts.

Click to expand...

No tone curve for X-Rite's software, at least as of 2-3 years ago. Just a basic dual-illuminant matrix. No LUTs (not even the "2.5D" LUTs used by Adobe and dcamprof), no tone curve.

Click to expand...

Interesting, I never used it but without a tone curve related output would typically look flat. Does it produce a dcp or ICC input profile that can be looked at?

Click to expand...

DCP - I haven't used it in ages because dcamprof was so superior.

I don't know how Adobe behaves when the DCP profile lacks a tone curve - maybe when it is completely missing (as opposed to present but indicating linear mapping) it defaults to something not linear?

RawTherapee will default to ignoring the built in tone curve (although currently defaults to using a HueSatMap entry that is tied to tone curve, I've filed an issue to work on making this more consistent after 5.9 gets out - https://github.com/Beep6581/RawTherapee/issues/6467 )

Click to expand...

Interesting! I have never heard of dcamprof until now. I actually had the thought that surely there must be some open source software more advanced than x-rites own. Can I hear more about your experience with the software? I have been frustrated that even with a 61MP capture of the Passport, their software will either fail or generate an awful profile if the passport doesn't take up the majority of the frame. Curious if this other software can do better.

dcamprof is actually the core of Lumariver, a command-line program containing all the essential functions. Both were written by Anders Torger, who also contributed to RawTherapee development. Thing is, while the program source code is readily available, there are no installers or distro packages of it. It's not that hard to compile on Linux distros, and really not that hard on Windows except it takes installation of something with a gcc compiler like msys2.

Even if you don't use it, its documentation is really, really good. I cut my teeth on color management reading it and Elle Stone's articles. Here are links to both:

http://rawtherapee.com/mirror/dcamprof/dcamprof.html
http://rawtherapee.com/mirror/dcamprof/camera-profiling.html
https://ninedegreesbelow.com/

And a link to the github repo hosting the dcamprof source code:
https://github.com/Beep6581/dcamprof

Colorchecker calibration, Gamma and Other correction techniques

Active member

Veteran Member

Senior Member

Veteran Member

Active member

Veteran Member

Senior Member

Veteran Member

Veteran Member

Forum Pro

Veteran Member

Veteran Member

Veteran Member

Forum Pro

Veteran Member

Forum Pro

Veteran Member

Veteran Member

Active member

Senior Member

Similar threads

Keyboard shortcuts