Did Adobe change something in the Adobe Standard profile for Olympus cameras with ACR8.3/DNGC8.3, which makes reds less saturated than previously and other camera models?
Here is the example side-by-side image from the Timbur Born dpreview thread post, linked above, with annotations added by me:
Here is the example photo from the original thread contributer, showing ACR 8.3 muting reds for many Olympus camera models, with yellow annotations added by me:
> TB wrote : ""
first iteration of E-M1 profile was just a copy of E-M5 profile (that's when Adobe first publicly released E-M1 __preliminary__ support), then there was a 2nd iteration when the camera apparently was profiled in the latest (the last one before final) 8.3RC and now there is a 3rd iteration in 8.3 final... so he is comparing apples to oranges... if you still have Adobe DNG converter (v8.2) - get the profile out (using Adobe DNG PE) and compare it w/ E-M5 profile using dcptool
This problem isn't only about the Olympus EM5 v the EM1. If you inspect the images of yarn above, for which the ACR gamut warning is clipping colors outside sRGB (I believe), you will notice the same problem is affecting the reds for the newest Nikon, Sony and Pentax cameras.
and that is done intentionally, from what I understand "Adobe" with recent profiles (Adobe Standard) is using LUTs ( ProfileHueSatMapData* ??? ) a lot for color transforms to prevent saturated colors clipping as much as possible for their mass user (and so they probably want to overdo the clamping to bring as much extremely saturated color as possible within ProphotoRGB coordinates and that probably means that lesser saturated /but still quite saturated/ colors will end up within AdoberRGB or even sRGB coordinates as a result, because you 'd assume the transform shall be smooth) ... w/ some previous profiles before you can remove ProfileHueSatMapData* tags (and ProfileLookTable* tags) from profile to make it "pure" matrix and still get a usable profile - no more (unfortunately), now these ProfileHueSatMapData* LUTs are not only a postfix for some colors after a matrix operations (to bring them within the Prophoto (?) colorspace coordinates), but they do a lot of color transform work... so if you like the clipping you shall lobby Adobe to provide as a separate download (not included in distribution for general public) a set of pure matrix profiles and live w/ clipping...
PS: alternatively go to camera calibration tab and increase the saturation of the proper primary (red) to your taste, save that as a preset (or do in DNG PE and save it as a profile)... that's just an inverse of what Adobe (EChan) was suggesting for reduction the oversaturation and clipping when it happens w/ existing profiles.
Hi rarewolf and folks,
To be clear: ACR is not "clipping" colors to the sRGB or Adobe RGB gamuts in the latest color profiles for the latest cameras.
It used to be the case that primary color correction in ACR (e.g., for previously supported models) was accomplished by color matrices, which has a side effect of clipping some saturated colors (bright flowers, colored light sources, etc.). This situation has now been improved with more recent color profiles, where saturation mapping is smoother to prevent clipping -- i.e., nonlinear compression of saturated colors, similar to the Vibrance control.
Don't worry, you can still capture extremely vivid objects beyond the gamuts of sRGB and Adobe RGB ...
> This situation has now been improved
the irony, I guess, is that some does not see that as an improvement, but rather as an encroaching on what the colors shall be... I'd assume that if a particular color (after a color transform) used to fall within Prophoto, but outside of Adobe RGB (or sRGB) and now because of "smoother" "clamping" falls within Adobe RGB (or sRGB) it bothers some... that is separate from subjective (or not) arguments whether E-M1 profile makes people faces too yellow or too red
That should comfort the crowd! ... whom you cannot blame for seeing something change significantly from one version to the next, and wonder what happened. In any case, it's good to hear Adobe's justification ...
PS ... thanks also for the Olympus profiles that came with this version!
First of all: Thanks to Eric for chimming in and explaining the situation. Are the new profiles LUT based instead of matrix based now (and would that include the new Olympus picture mode profiles being LUT based)?
This is quite a *fundamental* change that should have been properly documented. Thus it would have spared us users the time and effort trying to find out what's going on. I remember another case where I found a change only documented in the *press release*, but not in the release notes. So please keep an eye on your release notes, especially if something changes as dramatical as the standard color profile.
I am surprised to learn that the old profiles could cause reds to be mapped outside of the already large PhotoRGB?! Is this what you mean by "clipping" intense colors? Saturated reds have been problematic in the past, so this improvement is generally very welcome. You should have published it as Beta, though, because there are several issues with the current implementation that may need further fine-tuning.
Some of my reds turn to rust/orange now. This is not only a problem of getting a starting point where reds look plain odd/wrong. This is a problem of the HSL picker identifying the same patch of my red wall as being *more* orange with the new E-M1 profile vs. the old E-M1 RC and E-M5 profiles. The latter means that I tested both cameras and all three profiles, using the HSL picker on the same point of my wall to pull down saturation. When red saturation reaches -100 then orange saturation reaches -68 with the new E-M1 profile, but only -54 with E-M1 RC and -55 with E-M5 (using an E-M5 image).
Same thing with only changing the Orange slider, it affects my red wall more with the new E-M1 profile than it did with the RC E-M1 and with the E-M5 profiles. On the other hand I noticed that my ocre (rather yellow) wall is affected *less* by the Orange slider now. So which profile do I trust more about the color of my wall and HSL slider behavior?
Just to mention it, personally I never implied that red is clipped to the sRGB primary. While it was reasonable to speculate this as one source of the issue I just observed and reported that the former over 50% out of gamut reds of my demonstration image are now almost completely (over 90%) within sRGB's red primary. And even while I understand why the new non-linearly compressed approach makes sense I still have my reservation about how it works in practice. An intense red wall should be shown as red, not rusty with clear orange tint.
PS to deejjjaaaa: Accusing other users of ill intent and pointing fingers while they only try to make sense of their working tools doesn't really help the discussion much. And using the camera calibration section to increase the red primary is *not* a solution, because it considerably affects saturation of yellow and orange, too.
> Thus it would have spared us users the time and effort trying to find out what's going on.
actually curious users could witness that happening for a long time by reading discussions here
> You should have published it as Beta, though,
and that was done too... albeit for other cameras (not Olympus)
> PS to deejjjaaaa: Accusing
that's your rich imagination, never did - except to point out that using E-M5 profile applied to E-M1 raw file "AS IS" shall not be done as an argument for anything
> And using the camera calibration section to increase the red primary is *not* a solution, because it considerably affects saturation of yellow and orange, too.
when you don't have other means that will do.
> I am surprised to learn that the old profiles could cause reds to be mapped outside of the already large PhotoRGB?
not only reds, but blues as well for example
I would assume that all newer profiles are LUT-based, especially the camera-matching ones.
The ACR 8.2 profile for the EM-1 was preliminary. The ACR 8.3 profile for the EM-1 is not preliminary so consider it final, unless there is still something wrong with it. I missed that the EM-1 was preliminary, before, otherwise I probably wouldn’t have posted anything in the first place.
Yes, the older EM-1 profiles were different than that which shipped with 8.3 but is there anything you’re saying is wrong, now, or just that the new one is different, which of course it is.
Personally I don't have a problem with the new E-M1 profile being "different" and welcome the overall improvement and hopefully implementation of LUT profiles. But the currnent implementation makes Lightroom render my purely red wall more rusty-brown than before and even identifies it as being more orange when the HSL sliders are used. And by "before" I mean not only in comparison with the preliminary E-M1 profile, but also in comparison with an E-M5 camera shooting the same target.
And which profile is right about the amount of orange in my red and ocre walls, new or old? One of the benefits of Lightroom's HSL section is the inclusion of a dedicated orange slider. But for my example it seems that this slider is working worse than before, affecting reds stronger and ocre less.
Changing the red primary in the camera calibration section affects *all* tonalities within the magenta-yellow-red triangle. I wouldn't call this a good hint and viable solution just to increase saturation of reds.
Are you saying that the HSL slider seems to be operating on a different range of on-screen colors based on what camera is used, or is it merely that the colors that are displayed are different, so the area selected is different?
If you’re shooting the same subjects with both an EM-5 and an EM-1 (or any mixture of old-matrix-based and new-LUT-based profiles) that could cause color-matching problems since one is using the older type of profile.
Yes, the HSL sliders operate different. When I use the HSL picker and pull down red saturation to -100 on a spot of my red wall then orange gets pulled down along with it to -68 with the new profile and only -55 with the old ones.
When I pull down the orange saturation slider to -100 on its own then my red wall is slightly more affected while my ocre wall is slightly less affected compared to old profiles. So Lightroom's HSL section using the new profile sees my walls as different colors compared to using the old profiles. I did not try the magenta part of the magenta-yellow-red triangle yet, but wouldn't be surprised if it was behaving somewhat differently, too.
Since I am not earning money with these tools and generally didn't use HSL too often in the past this is not a real problem for myself, but I am still reporting my observations. And of course this doesn't make switchting between my E-M5 and E-M1 easier. My main grief still is that a very red wall comes out rusty/brownish with the new profile. This just doesn't seem right from a common sense point of view.
The “red” wall is more orange in the new profile so it makes sense that the color the targeted-adjustment-tool is selecting affects orange slider more. If the red wall was halfway between red and orange then it would pull down the orange slider as much as the red slider—both at -100, right?
You’re not describing anything different than the HSL sliders are “seeing” an oranger wall with the new profile, just like your eyes are seeing an oranger wall. Are you expecting the HSL sliders to work in pre-color-profile colorspace?
You are correct, I am describing the HSL slider "seeing" the red wall as being more orange and behaving accordingly and I am not expecting them to work differently. Instead I am pointing to a possible problem with how the new profile affects the HSL workflow. If the profile makes my red wall turn more orange, then I have to go to some length getting its *original* red back.
It's not as easy as changing the red primary, it's not as easy as using the HSL picker and increasing saturation of the area, it's not as easy as increasing saturation via single HSL slider! I have to carefully change colors via HSL hue sliders first, hoping to not mess too much with other areas of the image, then change saturation. All of this just to make an *originally* red wall become red again (and an ocre wall more orange again, since ocre is not just yellow).
Quite a bit of extra work involved now. As a professional I would likely just built my own LUT profile via appropiate color targets, but as a family man shooting his kids it seems a bit much to ask just for getting our home's red walls back.
> Changing the red primary in the camera calibration section affects *all* tonalities within the magenta-yellow-red triangle. I wouldn't call this a good hint and viable solution just to increase saturation of reds.
I have to repeat - "when you don't have other means that will do"... what in "when you don't have other means" is so difficult ?
> This just doesn't seem right from a common sense point of view
you shall supply the raw file in question to Eric Chan and ask to look into the matter - it is quite possible that they can change the profile to address how the non linear color transform works to save your red from being clamped into orange...
For your red brick wall, Adobe’s newer profiles are affecting the red highlights in a non-linear way, via a look-up-table (LUT), diminishing the reds (and other intense color channels) to keep things from clipping visually most of the time, much like the Highlights toning slider does for overall image brightness, or bright blue skies remain blue in other situations.
Because the intensity of the red is diminished but the other channels are not, the green continues to grow shifting the red towards yellow, making the wall more orange.
In this case you sound like you’d be happy with the reds continuing to become more intense in their original linear way and managing the intensity further along in the process via an output profile or LR controls.
As far as what to do that brightens only the most intense reds, to somewhat reverse what Adobe’s profile has done to the color of your wall:
The sliders in the Camera Calibration area affect things linearly so cannot reverse the LUT’s non-linear highlight compression.
Have you tried brightening just the red channel in the Tone Curve?
I don’t know enough about the workings of a camera profile to know if creating your own with a color-checker would remove Adobe’s LUT or if it would just build on top of that?
> I have to repeat - "when you don't have other means that will do"...
> what in "when you don't have other means" is so difficult ?
Is it really the case that this 8.3 version of ACR demoaicing for these newer cameras is final, and that we shall expect no new development? That is, do we not have an oppotunity here within this forum to report that this version is delivering inaccurate reds?
Although I cannot verify Timur's observations because I'm not using his camera (yet), he does show obviously different reds as a result of Lr 8.2 v 8.3 development with all paramterizations being equal. I really do not believe it is Adobe's intent ask of its users (as Timur states) a demand for extra work (relative to past versions)(?) However, I do believe Adobe put this forum here for reporting such observations, and I refuse to believe you wouldn't report the same before adding extra steps to your workflow.
That said, I do believe Timur should create similar examples with respect to how 8.3 affects skin tones, and report back as to his preference for 8.2 v 8.3.
> That is, do we not have an oppotunity here within this forum to report that this version is delivering inaccurate reds?
I think based on previous examples that you need to supply Adobe with raw files illustrating some particular problem and that is the mort proper way of lobbying for some changes...
> he does show obviously different reds as a result of Lr 8.2 v 8.3 development with all paramterizations being equal
as noted quite different profiles were used...
>"I think based on previous examples that you need to supply Adobe with raw files illustrating some particular problem and that is the mort proper way of lobbying for some changes..."
I agree ... along with the OOC JPEG ...