X-Rite ColorChecker Passport

[madmanchan]

Hi Joseph, if you send me the file I can take a look and see what's going wrong. If the file is large, try YouSendIt.com with a target email address of madmanchan2000@yahoo.com.

[josephchiang]

100% would be 255 (more or less).

Also, I see some purple hook goodies next to the patches and that might be the issue. The software may be expecting black.

Oh, OK, I'll give that a try (removing the hooks) and put something behind the Mini to hold it up.  I did try using the X-Rite Passport software and it was able to recognize every target, but couldn't build a profile because it said that at least one channel was clipped, probably due to overexposure.  Could that be happening with the DNG PE as well?  That means ETTR and exp to the right can't work, despite being able to bring the exposure back under clipping.

[digitaldog]

That means ETTR and exp to the right can't work, despite being able to bring the exposure back under clipping.

You probably do want to back off on the exposure and see how it flies and no, you’re not pigeon holed into shooting the target and using the subsequent profile with the identical exposure settings. Eric might be able to explain why (I suspect due to the processing order of the use of the profile and that of the Exposure slider).

[madmanchan]

Yes, overexposure is most likely the issue here. You generally do not want to be too aggressive with ETTR when doing the color work. As long as you're shooting at low ISO and do not hideously underexpose the image, noise should not affect the color profile quality.

Basically, when a channel is clipped, it becomes very unreliable for the purposes of color estimation (at least, for building a profile). Obviously for natural images you can sometimes get away with it, with various highlight recovery tools, but that's not good for profile building.

Quite frankly most of the time when I'm shooting profile targets I just use the evaluative/matrix metering in Av mode ...

[digitaldog]

Yes, overexposure is most likely the issue here. You generally do not want to be too aggressive with ETTR when doing the color work. As long as you're shooting at low ISO and do not hideously underexpose the image, noise should not affect the color profile quality.
Basically, when a channel is clipped, it becomes very unreliable for the purposes of color estimation (at least, for building a profile). Obviously for natural images you can sometimes get away with it, with various highlight recovery tools, but that's not good for profile building.

It sounds like ETTR, useful for reducing noise in shadows isn’t necessary for the capture of the target and anything close to over exposure, certainly any channel clipping, causes issues with profile generation.

I guess an interesting experiment would be, bracketing the capture of the target without clipping and building a group of profiles. But based on what you say above, would it be fair to say, the net results would be the same quality profile?

[keith_cooper]

I've now been using the Passport device for a while, and have found that the extra convenience of the hard case means I actually remember to chuck it in my bag more often :-)

I've had the appropriate gear for ICC profiling for some years, but rarely ever needed it for my work - I know it's of use in some situations, but all too often it's more trouble than it's worth (YMMV)

The DNG profiles I've done for a few jobs have definitely improved the balance of colours under some fluorescent lighting, but I should emphasise that I'm not often doing colour critical work - Looking good normally trumps absolute accuracy ;-)

Some more 'Passport' observations: ColorChecker Passport review

[neil snape]

The advantages for the PassPort are many and many yet to be discovered.

Each type of photography will benefit in different ways.

I have and continue to do studio tethered shooting directly into LR. In the past I made camera calibrations via Bruce Fraser and Eric Chan's instructions which built a sometimes accurate profile but was not reliable for varying light only the light the camera calibration was made in.

The Passport makes a usable reliable profile that works in many lighting situations , even multiple ISO ratings.

The Dual illuminate goes much further in building a multi environment profile for mixed lighting.

In studio the raw colours based on the Passport profile are the best starting point, as Canon defaults are not accurate or precise, never have been, until now with the Passport.

Yet , you see , the base profile from Passport is the key, then you add lights, filters whatever you want , and the base camera profile is the best starting point. Easy to use, extremely helpful , a must have for any image producers already on top of their CM system.

[josephchiang]

I was able to create several camera profiles (diff't studio lights & modifiers) for my studio shots of antiquities in the last several weeks, thanks to the generous help of Andrew and Eric,    ....  although I still found the necessity to WB despite having created and used a custom camera profile confusing.

In any case, I took several shots and the resulting color was already good with the standard profiles.  I then applied the custom profile and the image got a little darker and slightly more contrasty.  I still had to adjust the tint ever so slightly to get more of the green that I was seeing in the object under almost all lighting conditions, but failed to appear under studio lighting conditions.  I'm not sure if this is a result of an optical effect (physical interaction of studio lights partially transmitting thru translucent object) or something else.  Perhaps we need to purchase an Eizo CG301W.

Thanks to all for the help!    

[eronald]

I've now been using the Passport device for a while, and have found that the extra convenience of the hard case means I actually remember to chuck it in my bag more often :-)

Yes, I consulted for Xrite on the photographic aspects of this product, and I stressed the importance of robust packaging and in-use protection for a tool that needs to be taken to location and used there.

Edmund

[probep]

I am disappointed with ColorChecker Passport (CCP) in particular and DNG-profiling in general. I've tested CCP aganst ProfileMaker for various lighting conditions (D50, D65, A, mixed). This program has nothing with color accuracy. Color differences between CCP values and values measured by i1Pro (and taking into account the effect of metamerism for non D50 lighting conditions) sometimes exceeded 25 deltaE [76] for ColorChecker patches. (Color differences for ProfileMaker did not exceed 2-3 deltaE). And CCP has no settings at all.  
By the way, what is the purpose of the file "ColorChecker24_spectral.txt" in CCP directory "win\Reference"? I thought that it was the reference. But it is NOT used by CCP.

[neil snape]

I am disappointed with ColorChecker Passport (CCP) in particular and DNG-profiling in general. I've tested CCP aganst ProfileMaker for various lighting conditions (D50, D65, A, mixed). This program has nothing with color accuracy. Color differences between CCP values and values measured by i1Pro (and taking into account the effect of metamerism for non D50 lighting conditions) sometimes exceeded 25 deltaE [76] for ColorChecker patches. (Color differences for ProfileMaker did not exceed 2-3 deltaE). And CCP has no settings at all.  
By the way, what is the purpose of the file "ColorChecker24_spectral.txt" in CCP directory "win\Reference"? I thought that it was the reference. But it is NOT used by CCP.

That is a large difference.

Not sure I follow the procedure though.

The new chart is made to have a quite uniform spectral response so I can't see the point in having controls for bias of a presumed amount of variance in distribution from D50?

The idea is shoot the chart in the light you are capturing , the response of the captor to the scene or chart in this case have to include the spectral differences. In theory that DNG profile then should place the values so that correct metamerism is reflected in the output of the source colours to working space.

It's obvious it isn't working for you. I would like to find out more. As many are trying on different cameras, systems etc there may be times when it doesn't work. Perhaps you are one instance.

It is working fine on my Canon 5DMKII against any colours I have been shooting and testing visually. I am not going to try to analyse numerically what I don't need to as if the colours repro'd for me are what are sitting in front of me, it does what it should , for me, and my gear.

What I did when I first got it was place the Passport on different copy work, prints, different media, different light, and then built profiles. In each case the colours came in far more precise than using the stock profiles. Again , I know the contrast is boosted, and certain colours are too saturated, but others are so much better than the stock LR profiles.

I only have one Dslr, so I can't test others just yet.

That is why I am very interested in your set up, why it's not working. How many users is it working for , how many not so well, and how many not at all. Delta 25 is for me a too large of error, but again compared to what exactly?

[eronald]

I know the contrast is boosted, and certain colours are too saturated, but others are so much better than the stock LR profiles.

I only have one Dslr, so I can't test others just yet.

That is why I am very interested in your set up, why it's not working. How many users is it working for , how many not so well, and how many not at all. Delta 25 is for me a too large of error, but again compared to what exactly?

I got the effect of excessively saturated colors and contrast too (D3x, Tungsten), with the Passport software. This is definitely an issue.

I'll be delighted to meet up with you, Neil, if you wan to test more cameras.

Edmund

[probep]

neil snape
I own Canon 5D mark II and 40D only. I used an Eye-One Pro UV-cut spectrophotometer, standard X-Rite ColorChecker target and X-Rite/GretagMacbeth ProfileMaker v.5.0.8.
First I measured my X-Rite ColorChecker with i1Pro UV-cut and KeyWizard software.

D50 example

1. For shooting I used sunlight with D50 lighting condition. CRI was 97 (i1Pro and i1Share software were used).
2. The ColorChecker target was shooted as recomended. Then I built DNG profile in CCP software and got the image in ACR plugin.
3. ColorChecker patches of the image were measured in Photoshop v.10.0.1.
4. Differences between these values and true values were calculated.

Results are here.
Colors were oversaturated. The same was for D65, A and mixed lighting conditions.
BTW Adobe DNG Profile Editor is not better.

[neil snape]

neil snape
I own Canon 5D mark II and 40D only. I used an Eye-One Pro UV-cut spectrophotometer, standard X-Rite ColorChecker target and X-Rite/GretagMacbeth ProfileMaker v.5.0.8.
First I measured my X-Rite ColorChecker with i1Pro UV-cut and KeyWizard software.

D50 example

1. For shooting I used sunlight with D50 lighting condition. CRI was 97 (i1Pro and i1Share software were used).
2. The ColorChecker target was shooted as recomended. Then I built DNG profile in CCP software and got the image in ACR plugin.
3. ColorChecker patches of the image were measured in Photoshop v.10.0.1.
4. Differences between these values and true values were calculated.

Results are here.
Colors were oversaturated. The same was for D65, A and mixed lighting conditions.
BTW Adobe DNG Profile Editor is not better.

And what if you shoot a repro chart and print it. Compare it under normalised lighting. Does it really come out to your eye at >25dE?

If it does that would back up the measured values.

[bjanes]

neil snape
I own Canon 5D mark II and 40D only. I used an Eye-One Pro UV-cut spectrophotometer, standard X-Rite ColorChecker target and X-Rite/GretagMacbeth ProfileMaker v.5.0.8.
First I measured my X-Rite ColorChecker with i1Pro UV-cut and KeyWizard software.

D50 example

1. For shooting I used sunlight with D50 lighting condition. CRI was 97 (i1Pro and i1Share software were used).
2. The ColorChecker target was shooted as recomended. Then I built DNG profile in CCP software and got the image in ACR plugin.
3. ColorChecker patches of the image were measured in Photoshop v.10.0.1.
4. Differences between these values and true values were calculated.

Results are here.
Colors were oversaturated. The same was for D65, A and mixed lighting conditions.
BTW Adobe DNG Profile Editor is not better.

Your results are impressive. I posted some results earlier, also showing that the Passport and Adobe DNG editor profiles boost saturation. I think that this may be intentional, since many users prefer saturated colors. Delta E is a standard for color errors, but it may not be the best way to compare profiles, since it involves luminance and small changes in exposure may affect the results. Also, a change is hue is more objectionable than a change in chroma (saturation), and the latter is actually preferred in some instances. I think the way Imatest color check reports the results is preferable.

Bill

[probep]

Your results are impressive. I posted some results earlier, also showing that the Passport and Adobe DNG editor profiles boost saturation. I think that this may be intentional, since many users prefer saturated colors. Delta E is a standard for color errors, but it may not be the best way to compare profiles, since it involves luminance and small changes in exposure may affect the results. Also, a change is hue is more objectionable than a change in chroma (saturation), and the latter is actually preferred in some instances. I think the way Imatest color check reports the results is preferable.

Bill

Thanks for your comment and the links.
Yes, you are right: people like saturated colors
When I need to get accurate colors on an image, I build ICC input profile. In this case color errors are minimal.

[madmanchan]

If you are trying to measure output-referred values (e.g., in Photoshop) and compare them to published reference values, they will likely be far off. In general the auto profiles created by the DNG Profile Editor and X-Rite software do not adjust the tone curve. Therefore, using these profiles within CR/LR will result in using the default tone curve within CR/LR, which has a fair amount of brightening and contrast added. Also the Blacks setting of 5 (default value in CR/LR) will step on the shadows a bit and also result in punched up colors. You can use a linearized tone curve within CR/LR by setting Brightness, Contrast, and Blacks to zero, and setting Point Curve to linear. You can then adjust Exposure so that the gray values match up reasonably well with the published chart values, assuming reasonably uniform illumination on the chart. The result will be closer to published values, but if you try to use these settings on real images, they will not look so good ...

[probep]

madmanchan
Thank you very much for the clarification.

[neil snape]

If you are trying to measure output-referred values (e.g., in Photoshop) and compare them to published reference values, they will likely be far off. In general the auto profiles created by the DNG Profile Editor and X-Rite software do not adjust the tone curve. Therefore, using these profiles within CR/LR will result in using the default tone curve within CR/LR, which has a fair amount of brightening and contrast added. Also the Blacks setting of 5 (default value in CR/LR) will step on the shadows a bit and also result in punched up colors. You can use a linearized tone curve within CR/LR by setting Brightness, Contrast, and Blacks to zero, and setting Point Curve to linear. You can then adjust Exposure so that the gray values match up reasonably well with the published chart values, assuming reasonably uniform illumination on the chart. The result will be closer to published values, but if you try to use these settings on real images, they will not look so good ...

Thanks Eric.
You always say things much clearer than I ever could.

What you say is what I was thinking, and clearly describes the reasons for such discrepancies.
I didn't really understand the comparison procedure which resulted in the large dE differences and as I had asked what was done to get from A to B>

I do think it is going to continue to be a repeating question of scene recording passing through to working space via a camera calibration or base set up being called a profile. To add to that confusion the beta version even had an ICC camera profile export which is quite a bit different than a DNG, or camera calibration *profile*.

Problem with black clipping in LR/CR is with the profiles created they are saturating some colours more than others. By backing off the default black clip from 5 to less the colours that are saturated , yes they becomes less saturated but so do the colours that are close to being where they should be.

Is it possible to use the DNG editor to tweak the PassPort profiles?

[Czornyj]

Is it possible to use the DNG editor to tweak the PassPort profiles?

I alredy tried it - and it works.