Raw & Post Processing, Printing > Colour Management

Custom Camera Profiling: A Look at the QPcard 203 book

(1/7) > >>

samueljohnchia:
Introduction
My own exploration into the world of custom camera profiling led me to discover the QPcard 203 book. This is an updated version to the QPcard 202, which is more prone to surface reflections because of less matte color patches. The free software, QPcalibration, that builds the custom camera profiles, automatically recognizes the cards and uses the respective set of reference values for its internal calculations. Very cool. No fussing about with controls points like in the Adobe DNG Profile Editor (DNG PE). The instructions on QP's website are straight forward and easy to follow.

At the moment I have as well two mini ColorCheckers from my i1 Pro kits, and I briefly owned a ColorChecker Passport too. I have used the ColorCheckers very successfully for profiling my camera for naturally lit (daylight) photos. I am pleased with the results. However, I'm not going to let the QP pass by me, especially if there was a possibility of better results!

This press release by QP, claims that their product is the better (and more economical) alternative to the X-rite ColorChecker Passport. While it does mention that the X-rite software only corrects for 6 hues and the QP card has 32 corrections for hue and 4 for saturation, the free DNG PE's profiles contains 90 corrections for hues, and 25 corrections for saturation. For dual illuminant profiles there are two such tables, and they are linearly interpolated for white balance temperature settings in ACR/Lightroom between 2850 K and 6500 K. For various reasons I do not recommend the X-rite Passport software, but the DNG PE. This is what I used for my comparisons. The corrections I refer to are found in the HueSatDelta tables of the DNG camera profiles, which can be decompiled using Sandy's dcpTool.

So the QPcard has 35 color patches whist the ColorChecker has but 24, the former with seven neutrals, and the latter with 6. QP claims: "All the patches, even though they may look similar on our cards use slightly different base pigments. The result is that our products use a higher number of spectrally unique hues than any other comparable reference card." Will a greater number of "better formulated" color patches but a reduced resolution of the profile's LUT be able to produce a better result than the ColorChecker + DNG PE approach? And there are also other issues that will affect the final profile quality. QP designed the 203 with a white background so that the software can automatically correct for unevenly lit targets (within limits), and it also corrects somewhat for lens falloff at the edges. Very cool again. But I am not sure if the white backing will result in strong enough veiling flare and obscure the darkest gray patch slightly.

Another thing to note is the QPcalibration does not support dual illuminant profiles, because according to the same press release, "a dual illuminant profile will only correct accurately at the two points where you profiled, all intermediate color temperatures
will get the wrong correction". Tim Lookingbill proved that this was not the case when the two illuminants are spectrally smooth, like tungsten and daylight. One will get achieve a profile that is generally usable across a wider range of correlated color temperatures that way, with more accurate warm colors. One will have to build separate profiles for daylight and tungsten when doing things the QP way.

First Impressions
Firstly, thanks are in order. The efficient folks at QP shipped my card from Sweden to Singapore just 5 working days after my order. This is excellent. The card is simply but properly and durably packaged, with no exterior damage to the card (or the packaging for that matter) during shipment.

Then I discovered some unhappy things. Firstly, there is a substance on the front cover of the card, which is slightly sticky to the touch. Perhaps it is some glue residue from gluing the paper/card components of the book together. This is quite disappointing workmanship.

When I opened the booklet, the unevenness of the surface of the large gray patch for white balancing was immediately obvious. From the large bump where the tying ribbon is attached, to small dimples everywhere over the surface, it is very irregular. The card must be carefully lit with diffused lighting otherwise clicking on various spots in the gray patch with ACR/Lightroom's White Balance Tool results in different temperature and tint settings.



Moreover, the surface of the gray patch is marred with long grazing scratches. I don't know what caused it, but they remind me of similar scratches on inkjet prints that are handled poorly. It should not be there on a new product! Also the cutting of the paper frame is badly done with creases and the edges are frayed. There is also some glue residue.

Unfortunately, the color patches are not in good condition either. Firstly, the one of the cream colored patches has a spot of aqua paint, possibly a splash from the adjacent patch during (I presume) the silk screen printing of the patches. Also, there are two other splash spots of paint, one of the cream patch, and another of the aqua patch on the white paper card.



Other patches have dust or flecks of dirt embedded in the patches. Blowing it away with a lens blower did not help - they are definitely stuck on the surface. I know that the QPcalibration software reads 6400 points off the photographed target and calculates from 3200 points, throwing away half of the outliers, but I don't think the patches themselves should be in such poor condition when new.

What is even more disturbing is how the patches are all slightly reflective. This image, taken with grazing light to the card, shows the problem at its worst. Scratches on the magenta patch, as well as across the most saturated red and sunset orange patch are also obvious. Note the surface sparkle. I don't know what is causing this surface sparkle (its more sparkly than a sheen), but it definitely is not a smooth matte surface like the ColorCheckers I'm used to. It is almost as though the patches were scuffed after the paint had dried, perhaps from stacking the cards with no protecting interleaving sheets so the back of one card rubs the patches of another, or perhaps from doing that before the paint had fully dried.



I only have experience with one QPcard at the moment, and while certainly I cannot categorically say that all the cards are flawed like mine, I have never had any issues with ColorCheckers, (nor have I ever heard of such problems) and I am quite disappointed that this card is in such poor condition fresh out. If others out there have QP 203s too, please let me know about the condition you received them in.

The Card in Use
Another issue that bugs me is that the QPcard 203 book is basically a cardboard that is folded in the middle. It therefore has "fold memory" and cannot be easily opened flat like the Passport. The passport to me wins because of its ability to stand on its own

and be opened to multiple angles - past 180 degrees - which is an advantage when placed standing (like an upside down V shape) on its own in a reference scene to be profiled. The QP203 book can only be made to stand vertically on its own, (it can certainly be held up at other angles with other props, but that is a nuisance) which is not ideal for most lighting situations unless one can get an assistant to hold the card at a suitable angle. Ideally, one wants the illumination to hit the target at about 45 degrees incident and the camera to be pointing directly at the card.

If the QP is standing on its own, the chart cannot be opened flat, and there is a possibility that the white paper backing and the color patches will reflect onto the gray patch and throw of white balancing with it (the gray patch has produced varying results for me thus far), or light may be thrown from the gray patch on to the colors. By the way, the gray patch of the QP is not as spectrally neutral as the Passport, as measured by Robin Myers over 2 charts. His review of the Passport also points out one additional advantage, that because the Passport's white balancing gray patch is much lighter, one can get a better exposure (greater signal to noise) especially in dimmer lighting.



Profile Building Settings in QPcalibration
The settings in QPcalibration are few, but important. There are three of them which can affect the resulting profile - "Curve Set", "Smoothing" and "Gamut Matching".



"Smoothing" makes almost no difference for daylight profiles. QP recommends it in the event one experiences posterization effects when photographing under unusual illumination. It is off by default, and I used that setting.

"Curve Set" by default is set to normal. At normal, it is identical to the 96 point curve that is the base tone curve in the DNG PE, except for the lowest 18 points where it is a bit lighter. I tried the high and low options but it only decreased the accuracy of the profile.

"Gamut Matching" comes into play depending on what one chooses as their output working space in ACR/Lightroom. Since I use ProPhoto RGB, it is set to large. Medium and Small are for Adobe RGB and sRGB respectively.

Profiling Results
Some general observations. QP's profiles are generally slightly lighter in tone overall than the DNG PE's profiles. Also it tends towards a more green-blue cast, but that is correctable using the white balance sliders in ACR/Lightroom. I preferred the grayscale rendition of the DNG PE, which was slightly more neutral, even after proper white balancing for the differing profiles.

For colors, the QP has a tendency to darker reds and darker blues, while foilage greens are consistently lighter when compared to the DNG PE's profiles. The QP also tends to make yellows a bit too saturated and neon-like.

Here is a tif file with the shot of the QPcard and ColorChecker, with each profile on a separate layer, in ProPhoto RGB. I also included the reference patches for comparison. I found the reference values from QP's own website to be quite a bit less accurate than Robin Myer's measurements, which I used for these reference patches.

Something to note is the rather harsh transitions from reds to greens for the QP profile, precisely because of the tendency to lighten greens and darken reds. This example shows the issue to a moderate degree.



Another user on dpreview has a similar albeit more severe issue from purples to greens: http://forums.dpreview.com/forums/thread/3227175?page=2#forum-post-41900656

Conclusion
After making a number of comparisons, I find that the QP is sometimes more accurate for some colors than the ColorChecker, usually in the lighter colors. This could be due to the presence of light color patches on the QPcard that is not found on the ColorChecker. However, I found the ColorChecker to be overall more accurate than the QP for the majority of daylight photos. The rendition of blue skies, foilage and wild flowers was not just more accurate, but also more pleasing. I also had trouble totally eliminating the slight greenish cast that is present in the QP profile. Although the QP is marketed as a more economical product, at this point I think the Passport ColorChecker is better made, produces better results more often (with the DNG Profile Editor, not the X-rite software!), and is better value.

The world of camera profiling is fraught with controversy, probably more so than display or printer profiling, since final rendering of tone and color is not only influenced by settings in a photographer's raw converter of choice, but also because rendering of color is very much personal. As Eric Chan has mentioned before, “Due to technical limitations, camera profiles are necessarily imperfect, and hence each software must make tradeoffs in terms of which color characteristics it is optimizing for. Since different software programs tend to optimize for different things, it is natural that the resulting profiles will also behave somewhat differently.”

Tim Parkin, who conducted the much heralded IQ180 verses 8x10" film re-tests, recently wrote about camera color too. This article here, (needs paid subscription but you can see most of the best parts), expounds on the issue of sensor metamerism, and hints at variable imperfect band pass filters on the color filter of the Bayer array, resulting in poorer or better color accuracy for different sensors. This follow up article (needs free subscription to view), explores similar issues further, and also discusses the photographer's preference. Whether or not the final result is "accurate" is a tricky proposition, and while the subsequent paragraphs will reveal what I prefer, your own results with your own cameras, choice of raw software, how it renders colors and tones and your own preferences will determine your preferred approach.

I will try to post more comparisons over the next couple of days.

Vladimirovich:
put both targets in one shot,

build profile from QP, build profile from XRite

convert using QP, convert using XRite

calculate dE for XRite patches w/ QP profile
calculate dE for QP patches w/ XRite profile

which one will do better ?

mac_paolo:
I recently bought the very same card.

I agree with Samuel. Quality is quite good but manufacturing has some issues.
My patches don't have any drop of neighbor colors, yet I found a very long mark which runs through several patches.

Qpcard support told me it shouldn't be an issue performance wise.

It's clear that Xrite CC24 may be less precise but better built. Overall it feels somewhat homemade.
It's not a bring everywhere product like the Xrite CC Passport.

samueljohnchia:

--- Quote from: Vladimirovich on February 15, 2013, 11:31:46 am ---put both targets in one shot,

build profile from QP, build profile from XRite

convert using QP, convert using XRite

calculate dE for XRite patches w/ QP profile
calculate dE for QP patches w/ XRite profile

which one will do better ?

--- End quote ---

I already did that and posted a link to a tiff file, layers, in ProPhoto RGB.

For what it's worth, it was made using a rather badly damaged QPcard 203 book with color patches in poor condition. I measured the patches with an i 1 Pro and they were reasonable close to the reference  values that I have - somewhere between QP's own references and Robin Myer's. So that is ok. But lighting the target well so that the patches don't show their scuffed marks is hard. I think I got it pretty much spot on with that one. Unfortunately I think I didn't have ideal illumination - it was a cloudy sort of day and I shot it in the park and there is an overall green cast, probably from some trees nearby. I will want to redo things when I get a new QP and proper sunshine in the open.

Also, I did not bother to calculate the dE for the patches. If you wish to do so you may. I may try that when I have a better comparison, but again the QPcard profile is just inaccurate enough that it's not funny. Interestingly it does not have a light blue sky patch, and for some reason that results in a too dark rendition of blues. Same for reds. Greens are too light. There is oversaturation of the reds, and when I have blue shadows, the blue channel often gets clipped when the QP profile is used. A check of the raw histogram indicates that the blue channel was not originally clipped for those captures. Yellows also get pushed towards green and renders with a neon-like hue. Not so for the ColorChecker. This is true even for fluorescent light tests that I did.  The relationship of the lightness of each color hue is really out of proportion and it results in distorted rendering of images from my camera. The ColorChecker does much better in this regard.

By the way all this was with Adobe Camera Raw in process version 2010 with all the settings zeroed out, tone curve linear (very important as that distorts hue and saturatio,n different over ranges of tones, affecting interpretation of the profile), ProPhoto RGB and 16 bits. Proper white balancing after applying the respective profiles. Lots of folks doing comparisons with QP and the CC forget to do this, or process out into Adobe RGB, which will not reveal how the profiles handle colors well outside the gamut of the target itself - that aspect has to be extrapolated and is very much dependent on the genius of the algorithms in the profile building software. The X-rite Passport software fails at that.

samueljohnchia:

--- Quote from: mac_paolo on February 15, 2013, 12:28:18 pm ---I recently bought the very same card.

I agree with Samuel. Quality is quite good but manufacturing has some issues.
My patches don't have any drop of neighbor colors, yet I found a very long mark which runs through several patches.

Qpcard support told me it shouldn't be an issue performance wise.

It's clear that Xrite CC24 may be less precise but better built. Overall it feels somewhat homemade.
It's not a bring everywhere product like the Xrite CC Passport.

--- End quote ---

I just received an email from QP that they would be sending me a replacement. While I am grateful for that, I am saddened to hear that they told you it will not be an issue. The color patches have to be properly lit for the scratches not to appear in the final shot, (even then they may still be visible, like in my case) and while the software takes enough readings to ignore the outliers, it is no excuse that the patches should look like that. No one wants to buy a brand new car with the scratch on the paint work, although yes, "it shouldn't be an issue performance wise". I'm not so sure about color patches on a color target though. It certainly has been an issue performance wise for me. The gray patch for WB is unreliable as it is in my current damaged card.

I guess most folks here have the Passport so it should be okay. The Passport's gray patch for WB is a beautiful thing.

Navigation

[0] Message Index

[#] Next page