Pages: [1] 2 3 ... 5   Go Down

Author Topic: More Color Checker questions WRT accurate color  (Read 42368 times)

RDoc

  • Newbie
  • *
  • Offline Offline
  • Posts: 18
More Color Checker questions WRT accurate color
« on: November 23, 2011, 02:31:37 pm »

What I'm attempting to do is photograph paintings to produce accurate prints. As part of this I'm trying to get an accurate input profile.

I photographed the Color Checker Passport with the histogram centered in RAW using Adobe color, then converted it to a DNG with Photoshop Raw. Then with Adobe DNG Editor made some profiles using both Linear and Base Profile tone curves and no other corrections. I then used those profiles on the original RAW image of the Color Checker with no corrections. The tone cure is clearly incorrect, so I tweaked it and the exposure to get the gray patches to match the published values. Incidentally, it seems very strange to me that the DNG Editor doesn't at least have an option to match the tone curve to the color patches.

The resulting color patches are quite close in both Adobe and ProPhoto, but certainly not exact and I'm wondering why. It seems to me that if I correct the Color Patch photo with the photo itself, the results should exactly match the values. Shouldn't the software force the colors to match? If not what is it trying to do?

I did the same tests using the XRite software with similar but somewhat worse results.

Logged

madmanchan

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2115
    • Web
Re: More Color Checker questions WRT accurate color
« Reply #1 on: November 23, 2011, 03:25:26 pm »

Did you manually adjust the color corrections in the DNG Profile Editor?  (The auto Chart Wizard feature may not be using the same reference values as the ones you're using.)
Logged
Eric Chan

RDoc

  • Newbie
  • *
  • Offline Offline
  • Posts: 18
Re: More Color Checker questions WRT accurate color
« Reply #2 on: November 23, 2011, 04:23:24 pm »

Did you manually adjust the color corrections in the DNG Profile Editor?  (The auto Chart Wizard feature may not be using the same reference values as the ones you're using.)

The values I'm using are the ones at:
http://www.brucelindbloom.com/downloads/ColorCheckerSpreadsheets.zip

I only edited the exposure and tone curve in the Raw editor so the bottom row of grays match the published values. Unless I'm really missing something in the UI, I don't see how it would be possible to manually match all the points in the color checker because of the way the controls and value readouts work. I've never seen a set of values in the same space the DNG editor uses. Is there a set of its patch values available?

Also, is there an easy way to match the tone curve to the gray values in the Color Checker? I did it by manually adjusting a curve with about 6 points to get the values correct and still have a reasonably smooth curve, but that's not very rigorous.
Logged

keithrsmith

  • Full Member
  • ***
  • Offline Offline
  • Posts: 118
Re: More Color Checker questions WRT accurate color
« Reply #3 on: November 24, 2011, 05:46:57 am »

Have you tried the software from Xrite instead of the adobe one?

http://xritephoto.com/ph_product_overview.aspx?action=support&id=1257

Keith
Logged

RFPhotography

  • Guest
Re: More Color Checker questions WRT accurate color
« Reply #4 on: November 24, 2011, 10:14:17 am »

Lindbloom's reference numbers are different from XRite's, http://xritephoto.com/ph_product_overview.aspx?ID=1257&Action=Support&SupportID=5159&catid=28.

If I look at RGB values for various patches in ACR then export that image as a dng and open it in Adobe's DNG Profile Editor, the numbers are way off.  Is there something happening in the Adobe application that's causing the issue?
« Last Edit: November 24, 2011, 10:21:09 am by BobFisher »
Logged

Tim Lookingbill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2436
Re: More Color Checker questions WRT accurate color
« Reply #5 on: November 24, 2011, 09:17:32 pm »

Lindbloom's reference numbers are different from XRite's, http://xritephoto.com/ph_product_overview.aspx?ID=1257&Action=Support&SupportID=5159&catid=28.

If I look at RGB values for various patches in ACR then export that image as a dng and open it in Adobe's DNG Profile Editor, the numbers are way off.  Is there something happening in the Adobe application that's causing the issue?


The RGB/Lab numbers, if I understand you correctly, as read from the DNG Profile Editor are suppose to be way off because they are read from a linear reference space.

If you're using Mac OS X you can use its DigitalColor Meter set to Lab to get the numbers according to the DNG Profile Editor preview.

But to be honest I've never had any trouble making my $500 Pentax K100D DSLR give me exactly (I mean EXACTLY) what my eyes see editing in ACR if I happen to have the subject captured right next to my display or at least close by to compare against.

I've made several different DNG profiles with the CCchart shot under various lighting and then apply each different profile to one image not shot under those lights and there's barely a difference in color shifts.

I find the most difficult colors are the ones whose weird spectral reflectance makeup gives a color that's way off from the original where it's easier and faster to fix it with ACR's HSL panel.

Paint pigments may or may not exhibit these weird spectral colors that fool the camera's sensor, but I can assure you a CCchart derived profile will only get you close to what you see overall but it's not going to fix those kind of color errors.
Logged

bjanes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3387
Re: More Color Checker questions WRT accurate color
« Reply #6 on: November 25, 2011, 09:59:17 am »

Lindbloom's reference numbers are different from XRite's, http://xritephoto.com/ph_product_overview.aspx?ID=1257&Action=Support&SupportID=5159&catid=28.

If I look at RGB values for various patches in ACR then export that image as a dng and open it in Adobe's DNG Profile Editor, the numbers are way off.  Is there something happening in the Adobe application that's causing the issue?

Probably the best reference for RGB coordinates for the ColorChecker is that provided by Danny Pascale. Values from different sources may vary according to the illuminant (1976 MacBeth values were for Illuminant C), the target color space, whether the data are RGB or R'G'B' (the latter are gamma encoded), and the method used for chromatic adaption. The Bradford adaption is less accurate than the value obtained by using the spectral reflectance of the patches, the SPD of the illuminant, and the 2 degree standard observer when one is using D65 spaces. ProPhotoRGB is less problematic than Adobe RGB or sRGB, because the latter two are D65 and require chromatic adaption from the L*a*b D50 whereas ProPhoto is already in D50.

Danny gives the Delta E for various measurements of current ColorCheckers and spaces, and they are relatively small for charts produced after 2005. For practical work, the published values are sufficiently accurate and one does not need a spectrophotometer to measure the patches.

Regards,

Bill
Logged

bjanes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3387
Re: More Color Checker questions WRT accurate color
« Reply #7 on: November 25, 2011, 10:31:59 am »

I only edited the exposure and tone curve in the Raw editor so the bottom row of grays match the published values. Unless I'm really missing something in the UI, I don't see how it would be possible to manually match all the points in the color checker because of the way the controls and value readouts work. I've never seen a set of values in the same space the DNG editor uses. Is there a set of its patch values available?

Also, is there an easy way to match the tone curve to the gray values in the Color Checker? I did it by manually adjusting a curve with about 6 points to get the values correct and still have a reasonably smooth curve, but that's not very rigorous.

The DNG Profile Editor ignores the tone curve established in the raw converter (ACR I presume). It does use the white balance set by ACR and stored in the DNG file, but when calculating the color table, it uses its own white balance. This can be shown by setting a very wild WB in ACR and using a dramatic tone curve.

Here is the ACR screen capture with a wild WB and dramatic tone curve:



When one opens the DNG in the Profile Editor, the tone curve is ignored but the WB is preserved:



When the color tables are calculated, the WB is adjusted by the profile editor:



You may be going about your work in the wrong way. I think you might try to use the profile editor to get accurate colors and then establish a custom tone curve in ACR to render the neutral patches with the proper luminances. Hopefully, Eric can comment.

Regards,

Bill
Logged

sandymc

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 350
Re: More Color Checker questions WRT accurate color
« Reply #8 on: November 25, 2011, 12:45:00 pm »

The values I'm using are the ones at:
http://www.brucelindbloom.com/downloads/ColorCheckerSpreadsheets.zip

I only edited the exposure and tone curve in the Raw editor so the bottom row of grays match the published values. Unless I'm really missing something in the UI, I don't see how it would be possible to manually match all the points in the color checker because of the way the controls and value readouts work. I've never seen a set of values in the same space the DNG editor uses. Is there a set of its patch values available?

Also, is there an easy way to match the tone curve to the gray values in the Color Checker? I did it by manually adjusting a curve with about 6 points to get the values correct and still have a reasonably smooth curve, but that's not very rigorous.

I'm a bit surprised that you're seeing a major difference in tone curve. But anyway, I did a bunch of work on calibration of various raw developers versus a GM 24 chart a few years ago that might be useful to you: http://chromasoft.blogspot.com/2008/01/lightroom-aperture-and-capture-one-mini_24.html

Sandy
Logged

bjanes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3387
Re: More Color Checker questions WRT accurate color
« Reply #9 on: November 25, 2011, 01:27:52 pm »

I'm a bit surprised that you're seeing a major difference in tone curve. But anyway, I did a bunch of work on calibration of various raw developers versus a GM 24 chart a few years ago that might be useful to you: http://chromasoft.blogspot.com/2008/01/lightroom-aperture-and-capture-one-mini_24.html

Sandy

Sandy,

I seem to get the best match to the nominal color checker values by using a linear tone curve in ACR. Here is the result using a Passport generated profile for my camera using a linear tone curve. Since most tone curves have an inflection point near the midtones, I adjusted exposure to give the desired value for gray patch 3 of the color checker.



I looked at your referenced web page, but could only see page 1. Where are the rest?

Regards,

Bill
Logged

RDoc

  • Newbie
  • *
  • Offline Offline
  • Posts: 18
Re: More Color Checker questions WRT accurate color
« Reply #11 on: November 25, 2011, 05:19:24 pm »


You may be going about your work in the wrong way. I think you might try to use the profile editor to get accurate colors and then establish a custom tone curve in ACR to render the neutral patches with the proper luminances. Hopefully, Eric can comment.


Yes, as noted in the OP that's exactly what I was trying to do.

  • I photograph the passport image in even lighting, Adobe RGB color
  • open it in ACR
  • save it as a DNG
  • use the DNG editor to create a profile with no edits
  • apply that profile to the original image in ACR
  • create a tone cure in ACR to match the values from the XRite site
  • open the result in Lab mode in CS5 using Adobe RGB and look at the patches

Essentially I'm using the DNG editor to build a profile, then applying that profile back on the same image used to generate the profile which I would think would result in an accurate color result even if the tone curve was wrong.

The result is acceptable in most cases, delta E of around 3-6 compared to the XRite published values but several are over 10 off even if I force the L values to be the same as the published values.

I did try all this using the XRite software as well and the results were generally better although there were still errors > 9 in patches different than the ones with large errors from the DNG editor.

Overall it does look like the biggest errors may be the DNG editor using patch values that aren't the same as the XRite Passport but there is still some kind of residual error in both. Anyway, if the results in values are noticeably off, perhaps there needs to be a way to tell the DNG editor what target it's looking at.

Certainly for many uses, accuracy isn't really what's wanted, but that's not the case for some applications, and as it stands I don't see how one can generate an accurate profile with either tool. There's no editing in the XRite software, and because of the way the UI works in the DNG editor I don't see how it would be at all reasonable to correct all the patches by hand.
Logged

bjanes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3387
Re: More Color Checker questions WRT accurate color
« Reply #12 on: November 26, 2011, 09:58:07 am »

  • I photograph the passport image in even lighting, Adobe RGB color
  • open it in ACR
  • save it as a DNG
  • use the DNG editor to create a profile with no edits
  • apply that profile to the original image in ACR
  • create a tone cure in ACR to match the values from the XRite site
  • open the result in Lab mode in CS5 using Adobe RGB and look at the patches

That is what I have been doing as well, except it is best to use ProPhotoRGB as the working space, since Photoshop would have to use the Bradford transformation to convert from the D65 of Adobe RGB to the D50 of L*a*b. A bit depth of 16 would help to prevent rounding errors. According to Danny Pascale, the Bradford transform produces an average ΔE of 1.4.

Essentially I'm using the DNG editor to build a profile, then applying that profile back on the same image used to generate the profile which I would think would result in an accurate color result even if the tone curve was wrong.

The result is acceptable in most cases, delta E of around 3-6 compared to the XRite published values but several are over 10 off even if I force the L values to be the same as the published values.

Since luminance is included in ΔE, failure in controlling luminance will cause ΔE to increase. BTW, how are you calculating ΔE? The CIED2000 is said to be the most accurate: i.e. correlates with observed differences in color. The various formulas are discussed by Norman Koren in the Colorcheck documentation. I find that program to be invaluable in evaluating the accuracy of the profiles (the DigitalDog would cringe on reading this ;D). The CIED2000 usually gives smaller ΔEs and makes the data look better as well as better correlating with observed differences (which I have not checked).

Here is the best that I can do with my D3 and the DNG profiler. The results might have been better had I constructed a dual illuminant profile for my tests which used Solux illumination which is about 4700K. The Passport results are slightly worse.





I did try all this using the XRite software as well and the results were generally better although there were still errors > 9 in patches different than the ones with large errors from the DNG editor.

Overall it does look like the biggest errors may be the DNG editor using patch values that aren't the same as the XRite Passport but there is still some kind of residual error in both. Anyway, if the results in values are noticeably off, perhaps there needs to be a way to tell the DNG editor what target it's looking at.

Certainly for many uses, accuracy isn't really what's wanted, but that's not the case for some applications, and as it stands I don't see how one can generate an accurate profile with either tool. There's no editing in the XRite software, and because of the way the UI works in the DNG editor I don't see how it would be at all reasonable to correct all the patches by hand.

The Colorchecker probably is not introducing that much of an error unless it is old or improperly stored. XRite recommends replacing the chart every 2 years. Here are the ΔEs that Danny Pascale found in evaluating 20 charts.



One must realize that perfectly accurate color is not possible with current digital cameras, since the respond differently than the human visual system to the various colors. In constructing a 3x3 matrix, the profile maker tries to minimize error, perhaps using a least squares approach, or perhaps favoring accuracy in various memory colors such as blue sky, foliage, human skin, etc. A landscape profile would be different than one optimized for portraits.

It would be interesting to learn what other users were obtaining with the DNG profiler and the Passport software.

If you need accurate reproduction under controlled conditions, it might best to use the Colorchecker SG (which uses many more patches) and make an ICC profile. This would entail considerable expense and you would have to use a raw converter (such as Capture One) that uses ICC profiles.

Regards,

Bill
Logged

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Re: More Color Checker questions WRT accurate color
« Reply #13 on: November 26, 2011, 11:10:31 am »

The resulting color patches are quite close in both Adobe and ProPhoto, but certainly not exact and I'm wondering why.

You should also ask yourself, IF (big if) you got those 24 colors to be 'correct' (or what some like to call accurate, then tell you it can't be "perfectly" accurate), what would the effect be on all the other millions of solid colors that make up your painting.

If the painting is a Macbeth, you seem to be in decent shape <g>.
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

bjanes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3387
Re: More Color Checker questions WRT accurate color
« Reply #14 on: November 26, 2011, 12:40:33 pm »

You should also ask yourself, IF (big if) you got those 24 colors to be 'correct' (or what some like to call accurate, then tell you it can't be "perfectly" accurate), what would the effect be on all the other millions of solid colors that make up your painting.

Andrew, I don't know what is the source of your hangup with "accurate". Correct is a more general term when used as an adjective. Accuracy is defined  as follows: "In the fields of science, engineering, industry and statistics, the accuracy of a measurement system is the degree of closeness of measurements of a quantity to that quantity's actual (true) value." If I calculate a ΔE for the green patch of the ColorChecker as rendered by my camera with a given profile and raw converter, I am measuring accuracy by comparing the observed value with the correct value as per the x-rite spec or an actual measured value from a laboratory grade spectrophotometer. Do you agree? Accuracy implies a degree of error that can be quantified, whereas correct does not, at lease as I understand the definitions.

As an authority in color management, how would you recommend that the OP calibrate his camera to obtain correct results when photographing artwork? I understand that museum professionals tend to use multishot backs with high color rendering index illumination and calibrate according to laboratory standards. If one lacks such a setup, what is the best way to obtain the best results with a dSLR?

Regards,

Bill
Logged

madmanchan

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2115
    • Web
Re: More Color Checker questions WRT accurate color
« Reply #15 on: November 26, 2011, 12:55:06 pm »

The problem with using reference numbers is that they're based on an assumed illuminant (e.g., D50).  So even if you average a bunch of chart values, that still doesn't account for the illuminant.  It's very unlikely that you photographed under illuminant D50, since it doesn't actually exist!   ;)   So, you photographed under some other (real-world) illuminant, which results in different actual values for the chart.  Put another way, the spectral radiances coming off the chart patches and into the camera in your test shot are not the same as the ones used to derive the reference numbers that you see posted on various sites ...
Logged
Eric Chan

RDoc

  • Newbie
  • *
  • Offline Offline
  • Posts: 18
Re: More Color Checker questions WRT accurate color
« Reply #16 on: November 26, 2011, 02:28:23 pm »

The problem with using reference numbers is that they're based on an assumed illuminant (e.g., D50).  So even if you average a bunch of chart values, that still doesn't account for the illuminant.  It's very unlikely that you photographed under illuminant D50, since it doesn't actually exist!   ;)   So, you photographed under some other (real-world) illuminant, which results in different actual values for the chart.  Put another way, the spectral radiances coming off the chart patches and into the camera in your test shot are not the same as the ones used to derive the reference numbers that you see posted on various sites ...

Yes, I understand that, but what I'm still not getting is why creating a profile using a color chart image, then applying that profile to the same image doesn't result in the values the software thinks the chart patches have. Isn't the software essentially building a transform that converts the input image data to the ideal color patch data? If so, at least for the color components, I'd expect that if I apply that transform to the data used to build the transform I'd get the ideal values.

Perhaps what I'm seeing are the DNG editor's ideal values? Is there a published set of values for the patches that the DNG editor uses?
Logged

bjanes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3387
Re: More Color Checker questions WRT accurate color
« Reply #17 on: November 26, 2011, 04:49:20 pm »

The problem with using reference numbers is that they're based on an assumed illuminant (e.g., D50).  So even if you average a bunch of chart values, that still doesn't account for the illuminant.  It's very unlikely that you photographed under illuminant D50, since it doesn't actually exist!   ;)   So, you photographed under some other (real-world) illuminant, which results in different actual values for the chart.  Put another way, the spectral radiances coming off the chart patches and into the camera in your test shot are not the same as the ones used to derive the reference numbers that you see posted on various sites ...

Eric, quite true. But one can use chromatic adaption as discussed in Section 5 of Danny Pascale's RGB Coordinates of the Macbeth Color Checker. One can also use chromatic adaption to derive R'G'B' values for other illuminants. X-Rite publishes values for D50 CIE L*a*b and D65 sRGB. Danny and Bruce Lindbloom give values for multiple illuminants. The other factor is the actual illuminant used to take the picture. In my case, I use Solux 4700K bulbs which are not that far from D50 and rely on Camera Raw to perform a transform from the camera XYZ data using white balance and whatever else to ProPhotoRGB D50. In any case, one can compare the observed rendered values under the employed illuminant and compare them to the published or measured values of the chart and obtain a measure of the calibration of the system--illuminant, chart, camera, and raw converter. Your comments would be appreciated.

Of one had the SPD of the Solux bulbs and the spectral reflectance properties of the chart, one could calculate the D50 L*a*b values directly using the 2 degree standard observer data, but this seems to be overkill for most purposes.

Regards,

Bill
Logged

Tim Lookingbill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2436
Re: More Color Checker questions WRT accurate color
« Reply #18 on: November 26, 2011, 06:22:43 pm »

Yes, I understand that, but what I'm still not getting is why creating a profile using a color chart image, then applying that profile to the same image doesn't result in the values the software thinks the chart patches have. Isn't the software essentially building a transform that converts the input image data to the ideal color patch data? If so, at least for the color components, I'd expect that if I apply that transform to the data used to build the transform I'd get the ideal values.

Perhaps what I'm seeing are the DNG editor's ideal values? Is there a published set of values for the patches that the DNG editor uses?


The numbers for that chart are derived from a machine (spectrophotometer) measuring the spectral response of each color, gray and black patches under a given amount of light and color temperature illuminant as Eric indicated. What that spectro "saw" is by the numbers and doesn't take into account human's adaptive nature to brightness, contrast and saturation of colors to any given scene viewed.

This is much like the affects on perception when the lights are turned on in a darkened movie theatre where our eyes immediately see a loss in contrast and richness. A spectro if it were possible to measure from the movie screen would still see the CCchart color patches as being the same because the spectro's is using its own light source and not the movie theatre's.

The default ACR settings which establishes contrast, saturation and brightness tries to normalize the appearance as a human would perceive it, not as a spectro would. You will note when adding contrast editing an image in Photoshop or ACR/LR, saturation is greatly affected while an attempt is made by the tools in the software to maintain hue which is all the profile can possibly control due to the non-linear nature of how humans perceive any given scene along with the non-linear nature of camera sensor.

The image samples below I took with my Pentax K100D DSLR in an attempt to get the gray and black patch readouts to measure as close as possible to the published Lab numbers using a curve adjust. My illuminant and light source is direct sunlight for all shots. You can measure yourself how close I got. The last image shows the profile applied along with the settings that gave exact Lab numbers to a real scene captured under the same light intensity. Note the lack of contrast of the overall appearance of the surrounding scene. This is the same effect that happens when the lights are turned on in a darkened theatre.

I can tell you for sure my eyes saw that scene as much brighter and full of contrast than what's depicted.
« Last Edit: November 26, 2011, 06:27:44 pm by tlooknbill »
Logged

MarkM

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 428
    • Alaska Photographer Mark Meyer
Re: More Color Checker questions WRT accurate color
« Reply #19 on: November 26, 2011, 07:07:03 pm »

You should also ask yourself, IF (big if) you got those 24 colors to be 'correct' (or what some like to call accurate, then tell you it can't be "perfectly" accurate), what would the effect be on all the other millions of solid colors that make up your painting.

My impression is that if your profile can reproduce these 24 colors to match published values then in theory other colors would fall more or less into place within an acceptable range. Isn't that the whole point of jumping through the hoop of making custom DNG profiles? If the only advantage of custom profiles is that you can accurately reproduce a macbeth chart (and even that seems questionable at this point) why would anyone do it? Why not just use a grey card and go from there?

While I generally opt for pleasing color, in this case it's quite easy to talk about accurate color. You have two sets of patches with known LAB values that you are comparing. This should be the one place where talking about accurate color makes perfect sense.

Also, all the talk of different spectral distributions is a bit beside the point. We're not trying to match spectral reflectance—I think we'd all be happy with a metameric match.

One question that comes to mind: are DNG profiles matrix-based? If so, it's certainly understandable that it would not be able to bring all the patches into line. If it's got a lookup-table though, I would expect the individual patches to match almost by definition.
Logged
Pages: [1] 2 3 ... 5   Go Up