reposting : BasiCColor Input 5... now with DCP

[Redcrown]

What do you think of this:

I shot the MacBeth ColorChecker Classic with my Canon 5D3 in direct sun.
Sun angle about 45 degrees. Lighting as even as possible, no nearby colorcasts.
24-105 lens at 97mm, Colorchecker filling center 1/3 of frame.
From bracketed shots I picked the one that has a Lab-L value of 96 on the white square.
Set custom white balance on the middle gray square, ACR with Adobe Standard shows WB 5550/+14

Generated 4 raw conversions with ACR 9.2, no adjustments other than profile and WB for each.

Dropped a 5x5 colorsampler on the red squares.
Read the Lab values and entered these into the Delta-E calculator at:
http://www.brucelindbloom.com/index.html?ColorDifferenceCalc.html

Used the LAB references values for the ColorChecker Classic from:
http://xritephoto.com/ph_product_overview.aspx?ID=824&Action=Support&SupportID=5159

Got the following:

BasicColor Profile... LAB = 52/76/41  Delta-E = 11.4
DNGPE Profile........ LAB = 54/66/36  Delta-E = 12.4
DcamProf Profile..... LAB = 55/60/27  Delta-E = 13.4
AdobeStd Profile..... LAB = 54/66/37  Delta-E = 12.3

So, what does that mean?

Do the huge Delta-E values mean my whole process is bogus?
Is the one with the lowest Delta-E the best?
Are the differences in Delta-E even significant?

[AlterEgo]

my whole process is bogus?

what was the aim ? if you want to do something close to "reproduction" profile/conversion and get decent dE-whatever numbers then you for example need to use a special conversion in ACR matched with how profile was generated by a specific tool... starting with process 2010 for example (otherwise you /might, based on the scene/ need to somehow compensate possible hidden corrections /compression/ closer to clipping done by Adobe in process 2012), not to mention other sliders and then DCamProf must be used in a certain manner too... you might want to disclose your exact parameters otherwise what you posted is akin to somebody coming to the forum swinging that default raw conversion settings in 4 different raw converters gives different results...surprise, surprise... that is about dE-yours > 10

[digitaldog]

Got the following:

BasicColor Profile... LAB = 52/76/41  Delta-E = 11.4
DNGPE Profile........ LAB = 54/66/36  Delta-E = 12.4
DcamProf Profile..... LAB = 55/60/27  Delta-E = 13.4
AdobeStd Profile..... LAB = 54/66/37  Delta-E = 12.3

So, what does that mean?

It means the delta's among the group are not significantly different from one another. I don't know it means anything else since accuracy here is kind of undefined and perhaps not your goal (instead pleasing colors?). Do you visually prefer one over the other? The dE's are so small, do they appear virtually identical? Does this aid in whether or not you specifically wish to drop $500+ on another product?

[Iliah]

> Do the huge Delta-E values mean my whole process is bogus?

Hard to say not seeing the raw file of the target on a neutral grey background, not knowing the goal (repro vs. "pleasing"), and not knowing the profiling parameters.

[Bip]

Set custom white balance on the middle gray square, ACR with Adobe Standard shows WB 5550/+14

Generated 4 raw conversions with ACR 9.2, no adjustments other than profile and WB for each.

If I remember correctly, for DCP profile, it is not useful to the white balance
(I do a custom white balance with the camera or D50/D55)

Used the LAB references values for the ColorChecker Classic from:
http://xritephoto.com/ph_product_overview.aspx?ID=824&Action=Support&SupportID=5159

Your Lab datas are ok? in relation to your target? (the right solution is to measure  patchs)

[Redcrown]

Thanks to all, I really appreciate the feedback.

ALterEgos response was most helpful. My original numbers were with ACR Process 2011. I ran conversions again using Process 2010, using the defaults (Blacks 5, Brightness +50, Contrast +25). That gave much lower Delta-E values, but strangely a much higher difference on the BasicColor profile.

BasicColor Profile... Delta-E = 6.30
DNGPE Profile........ Delta-E = 4.17
DcamProf Profile..... Delta-E = 4.46
AdobeStd Profile..... Delta-E = 4.26

My goal, I think, is to have a profile that produces the most accurate colors, not necessarily the most pleasing. It seems more reasonable to me to get from accurate to pleasing via adjustments than to start with inaccurate. Plus, old age and an eye injury have lowered my trust in my visual judgement. So I'm moving more and more to evaluations "by-the-numbers". My goal here is also to simply advance my understanding of the process. I have no intention of spending $500+ on BasicColor, and can't imagine what it would take to justify that cost.

I understand that perfectly matching the LAB reference numbers of the ColorChecker is unreasonable. It's just not possible for the average shooter to match the exact same D50 lighting conditions and exposure. But it's starting to seem reasonable to use the difference between Delta-E values to help judge the quality of profiles compared to each other.

I've followed several debates about how best to shoot a ColorChecker target for use in profiling. Even lighting, avoiding glare, and avoiding color reflections are the center of most of those debates. But there is little discussion of the best exposure to use.

Most guides just say use ETTR and avoid clipping on either end. In one old discussion, Eric Chan said the white square should have a Lab-L value of 96. The MacBeth reference values show Lab-96. The BasicColor documentation hints at the same.

However, I've shot the CC24 with bracketing that yields Lab-L values between 89 and 98. When I feed those shots to DNGPE and Xrite software, I get significant differences in the resulting profiles. Basically, the higher the exposure the higher the saturation. Hues don't shift much, but saturation does. An L96 profile can easily push saturation beyond the limits of Adobe98, while an L89 profile leaves room to spare.

Now I notice that my DNGPE profiles made from a LAB-L 96 shot generate larger Delta-E values than a profile made from a Lab-L 92 shot. However, the BasicColor profiles I've generated from different exposures do not show a significant difference. They are virtually identical. A mystery to me.

[AlterEgo]

I ran conversions again using Process 2010, using the defaults (Blacks 5, Brightness +50, Contrast +25). That gave much lower Delta-E values, but strangely a much higher difference on the BasicColor profile.

BasiCColor injects a specific tone curve / that is applied after exposure correction by ACR/LR code / for "art repro" profile I 'd assume aiming to compensate process 2012 in ACR/LR, but they never tell about this in manual (they tell about compensation but not which specific ACR/LR UI parameters)... remove the curve manually... if your profile was "art repro" (one of 2 options that BasiCColor gives for DCP profiles)... plus their DCP profiles built off 24 patch target are matrix profiles - others do some forms of LUT (unless you instructed DCamProf not to do LUT) profiles... and cetainly LUT "helps" for the patches off which it was built in the first place

[AlterEgo]

I've followed several debates about how best to shoot a ColorChecker target for use in profiling. Even lighting, avoiding glare, and avoiding color reflections are the center of most of those debates. But there is little discussion of the best exposure to use.

greatest exposure you can afford w/o any raw channel hitting the area (near clipping) where your camera model might have non linear effects or clipping... but may be also watch exposure time (how long) - some cameras after some limit will use forced NR - not that it shall affect anything, but then who knows... of some cameras can have heat from long exposure making results too noisy for dark patches or patches where one channel under your light is weak still (imagine blue patch under a weak tungsten light) - more serious

[Peter_DL]

I ran conversions again using Process 2010, using the defaults (Blacks 5, Brightness +50, Contrast +25).

The PV2010 defaults also include the Medium Contrast Point Curve.

To quote Eric Chan from PV2010 times: >> if you want to get exactly the tone rendering of the profile, you need to use the CR/LR default tone controls, which are Brightness 50, Contrast 25, and Point Curve set to Medium Contrast. <<.

In this case, with a "pleasing" tone curve in place, it might make sense to limit the comparison of the different profiles vs target values to an analysis of the delta-Hue and delta-Saturation, thus to ignore the different tonal states.

Peter

--

[torger]

I understand the high price, it's narrow speciality software that can't sell in large volumes regardless price. It's aimed at professional photographers which do reproduction work.

Anyway the elephant in the room when it comes to camera profiling software is -- are you going to use it with a tone-curve or not? It only makes sense to do Delta E comparisons on profiles without tone curve.

Many profiler makers are designed to make reproduction style profiles without tone curve which they often do well, and then they just slap some random tone curve on top and pretends that doesn't change color appearance. It seems X-Rite does that. I don't know about Basiccolor.

DCamProf has it's neutral tone reproduction operator to deal with tone curves. That is about subjective perceptual modeling whose accuracy cannot be measured with an instrument but you must use your eyes. More about tone curves here: http://www.ludd.ltu.se/~torger/dcamprof.html#tone_curves

[Doug Gray]

What do you think of this:

I shot the MacBeth ColorChecker Classic with my Canon 5D3 in direct sun.
Sun angle about 45 degrees. Lighting as even as possible, no nearby colorcasts.
24-105 lens at 97mm, Colorchecker filling center 1/3 of frame.
From bracketed shots I picked the one that has a Lab-L value of 96 on the white square.
Set custom white balance on the middle gray square, ACR with Adobe Standard shows WB 5550/+14

Generated 4 raw conversions with ACR 9.2, no adjustments other than profile and WB for each.

Dropped a 5x5 colorsampler on the red squares.
Read the Lab values and entered these into the Delta-E calculator at:
http://www.brucelindbloom.com/index.html?ColorDifferenceCalc.html

Used the LAB references values for the ColorChecker Classic from:
http://xritephoto.com/ph_product_overview.aspx?ID=824&Action=Support&SupportID=5159

Got the following:

BasicColor Profile... LAB = 52/76/41  Delta-E = 11.4
DNGPE Profile........ LAB = 54/66/36  Delta-E = 12.4
DcamProf Profile..... LAB = 55/60/27  Delta-E = 13.4
AdobeStd Profile..... LAB = 54/66/37  Delta-E = 12.3

So, what does that mean?

Do the huge Delta-E values mean my whole process is bogus?
Is the one with the lowest Delta-E the best?
Are the differences in Delta-E even significant?

As others pointed out, this is to be expected. You won't get close Delta-E's with output referenced photography. Further, it's a desired effect. Normally.

This is because the conversion is to an output referenced image. High luminance is compressed, mid level contrast is increased.

To get Delta-E's that are close produces an image that is a better colorimetric match but almost always these are not attractive on either a monitor or printed.

An exception is when you do repro work, for instance taking a photo of a photo. There, colorimetric accuracy is highly desired else the printed repro image will look pretty bad compared to the original photo, painting, or artwork you are trying to reproduce. This requires "scene referenced imaging" and colorimetric printing, usually using Absolute colorimetric intent. Torger discusses what is required for that. It's a somewhat specialized area.

[AlterEgo]

This is because the conversion is to an output referenced image.

one of my earliest tests from 2015 with dcamprof circa v0.7 or so, with Sony A7 (mark i) from the big topic , dE* is dE2K... that was conversion to an output referenced image too... in ACR.

[Doug Gray]

one of my earliest tests from 2015 with dcamprof circa v0.7 or so, with Sony A7 (mark i) from the big topic , dE* is dE2K... that was conversion to an output referenced image too... in ACR.

Perhaps you define "output referenced" differently than most. That's quite a reasonable match for a scene-referenced image. Though it should be noted that the 3 patches highlighted with dE2k's just over 2 actually have Delta-E's of around 5. Saturated colors, except for certain transition regions, typically have much smaller dE2k's than Delta-E's. But again, that's actually pretty normal for an ACR processed image with linear settings. Great for repro. Bad for normal photography.

Dcamprof does a pretty good job of both linear, scene referenced work or output referenced work with many selectable options.

A largely linear tone curve is evident from the neutral patches. dE2k's on neutral patches are not compressed like highly saturated colors yet the neutral patches do not exhibit the S-Curve normal output references images have.

From Dcamprof:

A scene-referred camera profile simply means that the purpose of the profile is to correct the colors so the output represents a true linear colorimetric measurement of the original scene. In other words we want the XYZ values for the standard observer, or any reversible conversion thereof. That is what we in daily speak would call an accurate linear profile (where linear means “no tone curve”, we can still employ a LUT for non-linear correction), which DCamProf makes per default.

An output-referred camera profile should instead produce output that can be directly connected to a screen or printer ICC profile and produce a pleasing output for that media. As discussed, for cameras this means in practice that there should be some sort of tone-curve applied to get a pleasing midtone contrast and compressed highlights. In other words if the camera profile converts to XYZ space, those XYZ values should already have the curve applied and also any other subjective adjustments.

[AlterEgo]

Perhaps you define "output referenced" differently than most.

a matter of wording then between "output referenced __image__" (= raw conversion/color transfer applied, does not matter whether the profile & conversion were aimed at repro or at pleasant effect) and "output-referred camera __profile__"

[Doug Gray]

a matter of wording then between "output referenced __image__" (= raw conversion/color transfer applied, does not matter whether the profile & conversion were aimed at repro or at pleasant effect) and "output-referred camera __profile__"

Possibly, the better term would be colorimetric. In a sense, both RGB images intended for repro printing and those intended for pleasing rendering of a high dynamic range scene could be considered "output referred images." They are, after all, both output.

But there needs to be common ways to distinguish them. What are your descriptive preferences for the rendered images?

[digitaldog]

Just checked out a demo, it built a nice DNG profile from a Passport but I didn't see much difference between it and a profile I built using X-rite's DNG software last year from the same target. I used the 'photo' not repro mode so this is of course subjective. The GUI needs some work, but no question, it can build a good DNG profile. The X-rite DNG profile is on the bottom, slight difference in blue patch:

[howardm]

The XRite one looks like it's got a bit of haze/fog over most of the colors.

[digitaldog]

The XRite one looks like it's got a bit of haze/fog over most of the colors.

Sorry, I'm not really seeing that.
Working on more examples, stay tuned.

[digitaldog]

Two groups of images with both sets of profiles. These are screen captures directly from Lightroom. They ARE in display color space for my wide gamut (PA272W) so you MUST view them with color management! These are TIFFs (not JPEGs in sRGB). The X-rite profile is top for all examples, BasICColor bottom. No other differences in the images otherwise of course.


http://digitaldog.net/files/BasICCvsXrite1.tif


http://digitaldog.net/files/BasICCvsXrite2.tif


Comments on difference welcome.

[digitaldog]

One more example. Built a DNG profile using the ColorChecker SG which this product supports. So, more patches (more is better?).
Again, this is a TIFF in Display color space so must be viewed with color management product:


http://digitaldog.net/files/CCPPvsCCGP.tif


Passport (Macbeth) on top, SG on bottom.
Subjective but to my eye, I kind of prefer the Passport profile but the differences are tiny IMHO.