DNG Profile Editor question.
Hi, everyone. I hope this message is within the range of topics with discuss here. It’s not about Colorsync but since we discuss color technology here, I thought this may be relevant and within the scope of subjects covered by the forum. I have started to use the DNG Profile Editor to create a profile for my camera + light combination. All the tutorials I've read and/or watched so far aren't clear about this aspect of the calibration or profiling process. I have used the ColorChecker Classic and ColorChecker Passport (both have the same 24 color patches). Here is the problem I am having: After I create the profile using DMG profile editor and a photo of one of the targets above, I try to read the value of the color patches but the measurements are very different from those of the patches. I thought and expected that after the profiling process the measurements I took on screen using the color samplers in ACR would display the same values as those on the color charts. For instance, the White patch has a RGB value of R-243, G-243, B-242 on the ColorChecker Classic target but when I measure my image of it on screen the values are R-246, G-245, B-244. Not a big difference here. However, the Black patch has a value of R-52, G-52, B-52 on the ColorChecker Classic target but the measurements on the screen are R-72, G-73, B-74. Just to illustrate my point: Red on target: R-175, G-54, B-60 Red on screen: R-225, G-68, B-76 Green on target: R-70, G-148, B-73 Green on screen: R-151, G-219, B-119 Blue on target: R-56, G-61, B-150 Blue on screen: R-62, G-86, B-207 Shouldn't the measurements on screen duplicate those on the target now that the camera has been profiled ? In this case what am I doing wrong and how can I improve my calibration process so my measurements on screen are accurate and reflect real life measurements ? Thank you in advance for your help.
On Mar 14, 2016, at 10:13 AM, Joe <joemailgroups@gmail.com> wrote:
In this case what am I doing wrong and how can I improve my calibration process so my measurements on screen are accurate and reflect real life measurements ?
Adobe products are all about "pleasing" color. Colorimetric accuracy is next to impossible using Adobe products -- and DNG profiles especially. You can sorta kinda get somewhat in the ballpark maybe if you're very patient. And very lucky. But even just getting DNG to _not_ apply a "contrast enhancing" S-shaped tone curve can be an herculean task -- and that's before you get to all the mucking around it does with hue angles and what-not. (And, in your specific case, you might have an additional layer of confusion to contend with -- depending on what you're using to measure the colors, the display profile might or might not be applied before and / or after your measurements. That is, just simply getting accurate measurements out of an Adobe product can be an exercise in futility.) I would first recommend making sure you know what it is you're after -- what goal you have in mind. If it's "better" or "more pleasing" color in portraiture or landscape or sports or similar photography, start with the DNG profile and tweak it to your liking. Adobe's tools for that sort of thing are as good as they get. But if your goal is colorimetric accuracy for fine art reproduction or product photography or the like, I'd strongly recommend against using any Adobe products at all as part of your workflow. Instead, you'll need something such as Raw Photo Processor that both traces its heritage back to Dave Coffin's dcraw and takes color seriously. And the profiles that ship with Raw Photo Processor in particular are better than any you'll be able to make for yourself without a _lot_ of knowledge and expertise and a bit of equipment that you'll either have to make yourself (which can be done very cheaply) or buy for more than you probably spent on your camera. Cheers, b&
Hi, Ben. First I wish to thank you for your detailed and helpful reply to my questions. It’s been very helpful. In regards to your question I am concerned with colorimetric accuracy over visual appearance in my specific application. In other words I don’t have a problem if colors appear slightly off on screen as long as I get good readings. I must use the eyedropper in ACR and read colors with colorimetric accuracy so that the values for real life targets are the same I measure from the same color patches on screen. Adobe Photoshop and ACR are simple tools to use and almost anyone knows a little about them. I would prefer to stay with them as tools even if it means that I have to add another piece of equipment or software for this task in order to create proper and accurate profiles. I tried the Raw Photo Processor you suggested but found it a bit difficult to use and lacking some of the tools easily available in Photoshop and ACR. Is there any solution you can recommend ? Thank you.
On Mar 14, 2016, at 10:32 AM, Ben Goren <ben@trumpetpower.com> wrote:
On Mar 14, 2016, at 10:13 AM, Joe <joemailgroups@gmail.com> wrote:
In this case what am I doing wrong and how can I improve my calibration process so my measurements on screen are accurate and reflect real life measurements ?
Adobe products are all about "pleasing" color. Colorimetric accuracy is next to impossible using Adobe products -- and DNG profiles especially.
You can sorta kinda get somewhat in the ballpark maybe if you're very patient. And very lucky. But even just getting DNG to _not_ apply a "contrast enhancing" S-shaped tone curve can be an herculean task -- and that's before you get to all the mucking around it does with hue angles and what-not.
(And, in your specific case, you might have an additional layer of confusion to contend with -- depending on what you're using to measure the colors, the display profile might or might not be applied before and / or after your measurements. That is, just simply getting accurate measurements out of an Adobe product can be an exercise in futility.)
I would first recommend making sure you know what it is you're after -- what goal you have in mind. If it's "better" or "more pleasing" color in portraiture or landscape or sports or similar photography, start with the DNG profile and tweak it to your liking. Adobe's tools for that sort of thing are as good as they get.
But if your goal is colorimetric accuracy for fine art reproduction or product photography or the like, I'd strongly recommend against using any Adobe products at all as part of your workflow. Instead, you'll need something such as Raw Photo Processor that both traces its heritage back to Dave Coffin's dcraw and takes color seriously. And the profiles that ship with Raw Photo Processor in particular are better than any you'll be able to make for yourself without a _lot_ of knowledge and expertise and a bit of equipment that you'll either have to make yourself (which can be done very cheaply) or buy for more than you probably spent on your camera.
Cheers,
b&
On Mar 15, 2016, at 1:52 PM, Joe <joemailgroups@gmail.com> wrote:
I must use the eyedropper in ACR and read colors with colorimetric accuracy so that the values for real life targets are the same I measure from the same color patches on screen.
[...]
Is there any solution you can recommend ?
I'm afraid not. The tools designed for colorimetric accuracy aren't very user friendly. The user friendly tools have serious problems with colorimetric accuracy. If you want something easy and are willing to sacrifice accuracy, get a ColorChecker Passport and use the DNG profile creator program that ships with it. The Passport is an indispensable tool, the best chart you can take in the field and better than most charts designed for the studio, so it's worth the money itself even if you never use the software. But the software will improve your results in ACR. If you want accuracy and are willing to work at it, start with Raw Photo Processor and learn how to control lighting in the shoot and how to get your white balance set right and so on. You'd be very hard pressed to create better general-purpose camera profiles than the ones Iliah ships with RPP...but they're general-purpose profiles. When you get to the point that you know why you'd want a specific-purpose profile, that's the time to go down that rabbit hole -- but you'll still want RPP to develop the images. Sorry I don't have an easy answer. As the saying goes, "Good, fast, cheap: pick two." Cheers, b&
From: Ben Goren
On Mar 15, 2016, at 1:52 PM, Joe <joemailgroups@gmail.com> wrote:
I must use the eyedropper in ACR and read colors with colorimetric accuracy so that the values for real life targets are the same I measure from the same color patches on screen.
[...]
Is there any solution you can recommend ?
I'm afraid not. The tools designed for colorimetric accuracy aren't very user friendly. The user friendly tools have serious problems with colorimetric accuracy.
Colorimetric accuracy is not the only consideration for a camera profile. The end result for the majority of photographers are visually pleasing images. I've had limited success going the DNG route, and use standard ICC profiles for most work. Camera sensors are imperfect, the range of colors captured in profiling shot(s) limited, light sources neither characterized fully nor matching an assumed model, etc. And those are only first-order effects - further complications include whether the sensor satisfies the Luther condition (aka Maxwell-Ives criterion). My experience has been that camera profiles made where only colorimetric accuracy is the priority suffer when applied to actual captures. Years back we tried characterizing cameras using as wide a range of single-frequency light sources as possible. The resulting profiles were exceptionally accurate. Perfect for product shots of widgets but not much more. Most scenes containing color gradations ended up having visually jarring, ugly transitions. I'm largely I agreement with Ben. The basic premise behind DNG profiles is sound. Characterize a camera using two significantly different but well-behaved light sources interpolate or extrapolate from there based on the lighting in the actual shot. The range of colors and tones on a standard ColorChecker is nowhere sufficient for either building baseline profiles for a specific camera model or fine-tuning for an individual body. Cheers, Ethan
On Mar 16, 2016, at 9:09 AM, Ethan Hansen <ehansen@drycreekphoto.com> wrote:
Camera sensors are imperfect, the range of colors captured in profiling shot(s) limited, light sources neither characterized fully nor matching an assumed model, etc. And those are only first-order effects - further complications include whether the sensor satisfies the Luther condition (aka Maxwell-Ives criterion).
That's an accurate summary of the biggest challenges in camera profiling...but they're all addressable or not as bad as one might fear. One of the biggest considerations is the scene illuminant, and whether you want to faithfully reproduce its appearance or make the scene appear as if it had been shot under some other illuminant, presumably a standard such as D50. The latter especially requires knowing the scene illuminant...but you can either measure it directly with a spectrometer or do a surprisingly good job of estimating it if you've a good idea what it was. For example, if you know that the scene is outdoors or lit by something incandescent and you have a known sample in the image (such as a chart, but could be anything you've measured), you can use that to determine the actual color temperature and thus the actual illuminant. This, of course, also gets you correct white balance -- something not achievable with an eyedropper unless the sample is spectrally flat. It also gets you a perfectly normalized exposure. Colors near the spectral locus are going to be problematic, and you may do your profiling a disservice by attempting to force it to deal with them. On the other hand, no output device can even come close to representing them, either, so there's no practical point in even trying. Even if you succeeded, you're not going to see the colors on your monitor, let alone in a print. That leaves some sort of objective analysis as the only reason you might want something like that...and digital cameras are so far from being the right tool for that sort of thing it's not funny. (And, though sensors aren't perfect, they're shockingly good. Non-linearity comes almost entirely from optical artifacts, especially lens flare but sometimes also internal reflections in the camera's mirror box. Get the lighting and geometry right for reproduction work and none of that is a factor.)
Most scenes containing color gradations ended up having visually jarring, ugly transitions.
That's an indicator of a poorly-constructed profile. But good profiles, especially ones meaningfully encompassing a significant gamut, are difficult to construct.... Cheers, b&
participants (3)
-
Ben Goren
-
Ethan Hansen
-
Joe