Custom Camera Profiling: A Look at the QPcard 203 book

[mac_paolo]

Another thing that I mentioned previously, that the QP generates profiles that are greener than the CC+DNG PE. I now know why. The gray patches in the QPcard are redder than the CC, both measurably and visibly. The QP software must be basing the color mapping after white balancing off one of the lighter gray patches, and that is skewing the profile towards green. If ones wishes to test a QP profile against a DNG PE profile, different WB values must be used for each profile.

Agree.
I keep the WhiBal shot for reference. My Adobe Standard, CC24 and Qpcard shots exhibit very different WB values.
That's annoying but unavoidable.

[solarj]

Unless the software let me see each color patch's lab value, I won't trust the generated profile. I don't trust my eyes, I trust RGB valus 

And I think it is a must for such kind of software to have this ability, but sadly none for xrite/dng pe

DNG PE provided some strange readings and those value do not change with the adjustment, totally useless, maybe QPcard can improve in this matter

[samueljohnchia]

Unless the software let me see each color patch's lab value, I won't trust the generated profile. I don't trust my eyes, I trust RGB valus 

And I think it is a must for such kind of software to have this ability, but sadly none for xrite/dng pe

DNG PE provided some strange readings and those value do not change with the adjustment, totally useless, maybe QPcard can improve in this matter

While I certainly agree that the color readouts from the photographed target should be reasonably close to the reference values, (if they don't come close then what on earth is the profiling software basing the mapping on?) I also don't place all my trust on RGB values. When I get very exacting matchings with the CC+DNG PE, I notice that it renders light greens and yellows way too dark - clearly obvious when used on a qpcard photograph. Of course the qp is not perfect on the CC either. Something's got to give to force those few patches into order.

The current solutions are designed to be idiot proof. You are not going to be able to see lab values in QPcalibration. You cannot even name profiles more than 16 characters - a limitation I find very irritating when I have to specifically label profiles with enough info when doing testing. You still have to build the profile and then test it outside of the profile building software. I cannot know if QPcard can improve in getting better color matching, but it does seem to be less sensitive in the blue changes - a good thing.

[solarj]

While I certainly agree that the color readouts from the photographed target should be reasonably close to the reference values, (if they don't come close then what on earth is the profiling software basing the mapping on?) I also don't place all my trust on RGB values. When I get very exacting matchings with the CC+DNG PE, I notice that it renders light greens and yellows way too dark - clearly obvious when used on a qpcard photograph. Of course the qp is not perfect on the CC either. Something's got to give to force those few patches into order.

The current solutions are designed to be idiot proof. You are not going to be able to see lab values in QPcalibration. You cannot even name profiles more than 16 characters - a limitation I find very irritating when I have to specifically label profiles with enough info when doing testing. You still have to build the profile and then test it outside of the profile building software. I cannot know if QPcard can improve in getting better color matching, but it does seem to be less sensitive in the blue changes - a good thing.

On further reading of their software, you might be able to buy a plugin and then be able to see the lab value, still not very direct, but better than nothing

And if the values are all perfect matching but you still feel the color is not right, then the problem is more likely on the monitor calibration. In principle, on a perfect calibrated monitor, if you generate a digital color patch with those reference values given by color checker manufacturer, you will feel it looks exactly like what you see with your eyes under daylight (of course the brightness of monitor can affect that a lot, so a very bright HDTV could be considered)

[digitaldog]

Unless the software let me see each color patch's lab value, I won't trust the generated profile. I don't trust my eyes, I trust RGB valus 

The only way to get those Lab values IS from RGB (in this case). If you have a tagged RGB document, then crop and size it down (nearest Neighbor) such that only 1 pixel represents each patch, I'd be happy to feed that into ColorThink Pro which will give you the Lab values FROM the tagged RGB capture. Or if you have ColorThink Pro, pretty easy to do this yourself.

[samueljohnchia]

you might be able to buy a plugin and then be able to see the lab value, still not very direct, but better than nothing

Which plug-in are you referring to? Is it QPcalibration inspect?

[samueljohnchia]

And if the values are all perfect matching but you still feel the color is not right, then the problem is more likely on the monitor calibration. In principle, on a perfect calibrated monitor, if you generate a digital color patch with those reference values given by color checker manufacturer, you will feel it looks exactly like what you see with your eyes under daylight (of course the brightness of monitor can affect that a lot, so a very bright HDTV could be considered)

There are more colors in the real world than just the patches in the ColorChecker. While a custom profile or Adobe's Standard can get you perfect or near-perfect matching between a shot of the CC and a synthetic target, it does not mean that other colors in the real world will also be reproduced correctly. Mentioned many times already is the difficulty of reproducing colors in blue flowers. Even trying to reproduce exacting colors in another target, like the QPcard, is impossible as I have proven. This is absolutely nothing to do with monitor calibration.

[solarj]

There are more colors in the real world than just the patches in the ColorChecker. While a custom profile or Adobe's Standard can get you perfect or near-perfect matching between a shot of the CC and a synthetic target, it does not mean that other colors in the real world will also be reproduced correctly. Mentioned many times already is the difficulty of reproducing colors in blue flowers. Even trying to reproduce exacting colors in another target, like the QPcard, is impossible as I have proven. This is absolutely nothing to do with monitor calibration.

Of course the ability to reproduce color depends on the display, an UHDTV can provide a much larger gamut, but since the data format on internet are mostly based on sRGB color space, that added capacity will not be utilized at mass scale until many years later

About calibration, if the characteristic of the sensor is enough good (one single matrix is enough to transfer to xyz color space with little error), calibrating red/green/blue patch to make them match the reference value will make the whole gamut matching more or less correctly (yellow/magenta/cyan might be less accurate)

[samueljohnchia]

Of course the ability to reproduce color depends on the display, an UHDTV can provide a much larger gamut, but since the data format on internet are mostly based on sRGB color space, that added capacity will not be utilized at mass scale until many years later

About calibration, if the characteristic of the sensor is enough good (one single matrix is enough to transfer to xyz color space with little error), calibrating red/green/blue patch to make them match the reference value will make the whole gamut matching more or less correctly (yellow/magenta/cyan might be less accurate)

Yes, for visual comparisons on a display, it will be limited by its gamut, but also by its spectral characteristics and precision of the entire display pipeline.

I think the question is, can a camera profile really allow one to map all colors captured by a camera system accurately? As Ernst and Eric Chan pointed out before, it is not possible. There are always compromises one has to make for a camera profile. Your idea about matching just R, G and B is somewhat similar to Dunthorn's Calibration. My own limited experience with ColorPerfect has shown me that it does produce very good results too. Less twisting around with complex LUT tables.

[mirekti]

Hi all, has there been any new updates on this product? I've been using ColorChecker for some time, and I would like to try this tool as the DNG PE results are not so great. Much better than X Rite, though.

[mirekti]

I received QPCard 203 in the mean time, and already made few profiles. I am not sure if I do something wrong, but it seems my photos have a slight green cast to it (I solve this by reducing the temp by 200-300), and blue is a bit darker than expected.
Is there a proper way to create let's say a daylight profile?
I can see one can download reference files from here http://www.qpcard.com/en_b2c/colors-on-card What should I used them for? There is a tiff file with the target, and txt file with bunch of numbers. Should I use them in some way to compare it with the target I shot? I really apologize for my ignorance.

[Ligament]

On Sept 12 I sent a request for a replacement card to qpcard, as mine had delaminated.

On Oct 22, I got a reply they would issue a replacement, which I am grateful for.

So they are still in business, but a delay of 40 days from inquiry to response is truly horrid and therefore I would suggest to photographers who rely on qpcard on a professional level to be very wary of this product if your job and reputation depends on it.

[ErikKaffehr]

Hi,

My results with different options:

http://echophoto.dnsalias.net/ekr/index.php/photoarticles/79-p45-colour-rendition?start=5
http://echophoto.dnsalias.net/ekr/index.php/photoarticles/79-p45-colour-rendition?start=4


Best regards
Erik

I received QPCard 203 in the mean time, and already made few profiles. I am not sure if I do something wrong, but it seems my photos have a slight green cast to it (I solve this by reducing the temp by 200-300), and blue is a bit darker than expected.
Is there a proper way to create let's say a daylight profile?
I can see one can download reference files from here http://www.qpcard.com/en_b2c/colors-on-card What should I used them for? There is a tiff file with the target, and txt file with bunch of numbers. Should I use them in some way to compare it with the target I shot? I really apologize for my ignorance.

[mirekti]

Today I tried to create a new color profile and it was made at 2.30 PM under direct sunlight.
What I don't understand is the correction table. I can see old camera values, target values, and new values. Of course not all new values don't match the target values.
Any idea how I could fix this? What causes this problem? Here is an example.