Achieving good color in raw conversion, what is your view?

[ErikKaffehr]

Hi,

We had a lot of discussions recently about color rendition in different raw converters, much of the discussion being about color rendition of the Nikon D800 which seems not be to everyones liking. Some users suggested the need of three different raw converters for different purposes.

Personally, I have little issues with camera color, but I don't shoot portraits or commercial work, I see that needs are different.

I got the impression that both Michael Reichmann and Jeff Schewe feel that a good raw converter like the one in LR 4.2 and ACR can be tweaked to give a satisfactory color whatever the need.

My reasoning may be that it may be better to spend some time tweaking a single raw converter to your taste than live with several ones.

For me, a reasonable approach would be to "zero out" the camera/raw processing pipeline by creating a DNG profile using the DNG Profile Editor from Adobe or Xrite's Color Checker Passport.

DNG Profiles can be tweaked in DNG Profile Editor. In my humble opinion the smartest approach would be to just create some preset for different kind of color rendition.

What is your take on the issue?

[Steve Weldon]

I've always taken it for granted that I'd profile each body for general use, and occasionally for specific use.. especially where heavy greens and red come into play.   

And I've always missed my Mange Nilsen profiles.  I still have and use them for my 1ds2.  But I  miss not having them for my 5d's, 1ds3's, etc..

Color is all about the eye, and more than one color takes more brain bits, more than two even more brain bits, etc, etc,, until lately I'm running a 14bit brain.  Unfortunately it's becoming increasingly less compatible with my 4 bit body.   

[stamper]

Colours should be processed to suit the vision of the photographer who took the image - unless working for a client - and I don't see it as a complicated - in theory - issue. Use the converter that suits your vision. I agree with the zerod out theory. Start with a neutral setting in the converter and processing to taste. As a test take an image you like and process it from as a neutral position as possible. Leave it for a week. Start over from scratch a week later and process it again. Repeat a couple of times if you wish and look at the results side by side. They will have differences in the final output. Now which is the correct one with regards to "accuracy"? None of them. They all will be pleasing rather than "accurate."

[smthopr]

I'm not sure I can help, but a little anecdote:

I have a Canon 5d.  Color is attractive with RAW processed in Adobe or Canon software.  Very easy to get pleasing color.

I've shot a couple movies in RAW mode with a certain "popular" digital cinema camera.  Color rarely looks pleasing coming out of the RAW converter. (there is only one RAW converter available).  After RAW conversion, the images need a lot of work to look good. But they never really look the way they did to my naked eye.  They can be good, but different.

An example:  A woman in a bright red dress.  When the skin is pleasing, the dress is dull with too much cyan in the dress.  Yes, one can select the color range of the dress and correct it, but it's a pain.

Another:  Two white men in the same frame, in the same lighting.  One looks normal, the other like a lobster.  Why?

It's either/or both the RAW conversion math.  Or, the colored filters on the chip are not pure enough.  I don't know the answer here. Could be a combination.  Perhaps in an effort to boost ISO sensitivity, weak colors filters are used on the chip that can't be decoded accurately in the RAW converter.

I don't know if the above applies to your Nikon D800, but my point is not to assume that because something is recorded in RAW mode that any correction is possible if you have the skill.  Sometimes the hardware/software makes an important contribution. Not all RAW data is equal.

[Anders_HK]

For me, a reasonable approach would be to "zero out" ...

When I years ago shot dslr I shot Nikon D200. One of things that frustrated me most was that the colors were not at a "zero out" stage (not correct and not pleasing colors when I opened at defaults). Instead they needed to be tweaked to first bring towards "zero out", prior to that I was comfortable with making my own adjustments. Ok, that was some five years ago. I have since shot MFDB and have not had the same problem because already at defaults, in particular in Capture One Pro, there is a "zero out" as I open at defaults and upon for me to make adjustments.

Simply, I feel any camera at this time should already be able to provide "zero out" for us to have a decent (pleasing) departure point to commence adjustments from. If like some have reported there are reddish hues on skin tones from D800 there sounds something wrong, either with profiles in the RAW converter, or perhaps in the processing chain in the camera? Is it perhaps that the D800 is so poor calibrated or profiled from factory, or simply that Nikon has not work with others to distribute their profiles etc?

Regarding landscapes vs. portraits. Actually I used to experience more difficult to arrive at pleasing rendition of landscapes than portraits, and it can still be so. There are more colors in landscapes simply, while for portraits the important is the skin tones. Best converter I have used for adjusting skin tones and colors and more is Capture One Pro.

Of course, one could calibrate any camera to a color checker in Lightroom or Camera RAW, but... then I guess the camera would adhere to Color Checker standard, which we may or may not find as pleasing. For my current Leaf back I would not dream of doing so because I find the colors very pleasing at my defaults in Capture One Pro. Colors are critical for an image, and it would seem that ten years into making of dslrs such D800 reddish hues or Canon greens should be things from past?

For someone using Capture One Pro for e.g. D800 and having problems, it would appear logical to open a report case to get help towards getting the issue resolved from Phase One, or at least narrowed down to if in fact it is the camera that is problem and thereby request Nikon to fix it?

Best regards,
Anders

[deejjjaaaa]

as for C1 one needs to check if D800 profile is in fact D800 profile (and not something else)... couple of years ago PhaseOne released C1 with profile for cameras like Pentax K5 (Sony 16mp) simply copied 1:1 from Pentax K7 (Samsung 14mp sensor)... and not only for Pentax K5 - they did the same for some other cameras (Panasonic GH2, GF1, etc).

[Hening Bettermann]

What is your take on the issue?

I am only interested in one color/profile: the "neutral" or "natural" one - as accurate as possible within the technical limits of camera, screen, print, software. To this end, I have made an ICC profile for my camera using the ColorChecker and Argyll. Then I process in Raw Developer with this profile as input and output profile, everything else zeroed. Later, in my pixel editor (PhotoLine) I "normalize" to Lab. (Convert to ProPhoto for print only; to sRGB for web). As far as I can see, this is the most "accurate" color I can muster. Since I shoot landscape in daylight exclusively, I feel I can get away with just one profile.

I find that the white balance is a parameter of gross impact, but that is a topic of its own.

In my next go at improving my workflow, I look forward to  trying the QP target and software:

QPcard 202  http://www.qpcard.com/en_b2c/color-reference-cards.html

QPcalibration ICC  http://www.qpcard.com/en_b2c/applications/qpcalibration-icc-profile.html

I also look forward to explore CIECAM02.

[deejjjaaaa]

I am only interested in one color/profile: the "neutral" or "natural" one - as accurate as possible within the technical limits of camera, screen, print, software. To this end, I have made an ICC profile for my camera using the ColorChecker and Argyll. Then I process in Raw Developer with this profile as input and output profile, everything else zeroed. Later, in my pixel editor (PhotoLine) I "normalize" to Lab. (Convert to ProPhoto for print only; to sRGB for web). As far as I can see, this is the most "accurate" color I can muster. Since I shoot landscape in daylight exclusively, I feel I can get away with just one profile.

I find that the white balance is a parameter of gross impact, but that is a topic of its own.

In my next go at improving my workflow, I look forward to  trying the QP target and software:

QPcard 202  http://www.qpcard.com/en_b2c/color-reference-cards.html

QPcalibration ICC  http://www.qpcard.com/en_b2c/applications/qpcalibration-icc-profile.html

I also look forward to explore CIECAM02.

you might want to consider RPP ( www.raw-photo-processor.com/ ) - it is quite suitable for what you do...

latest beta = http://www.raw-photo-processor.com/RPP/RPP64_1608Beta.zip

you can output to Lab, camera RGB w/o any color transforms (16 and 32 bit), Beta RGB (16 and 32 bit), etc

create your own profiles with embedded Argyll from xrite, qpcard and other targets natually or use embedded camera profiles prepared by Iliah Borg.

[ErikKaffehr]

Hi Anders,

The idea with the color checker is to reproduce colors correctly. So, there is no color checker standard. The color checker is simply a card with 16 very well defined colors. The colors are chosen to be somewhat relevant photographically but also to have good metameric properties, what I mean be relatively consistent under different types of illumination.

So you take a picture of the color checker under two illuminations, daylight and incandescent (halogen for instance) and generate a profile that reproduces the colors as well as possible. So, in the ideal case, the color checker would reproduce exactly. Such a profile would be correct, but not necessarily pleasant. What I have seen generally that raw converters exaggerate saturation a bit.

Lars Kjellberg has produced another standard card called QPCard with accompanying software. The QPCard has a different set of color patches that may possibly more suited for portraiture.

Anyway, what I would suggest is to use either Color Checker or QPCard to generate a calibrated color and tweak the colors for pleasantness.

Another post indicated that Capture One is quite sloppy with camera profiles, it seems that they just copied and renamed camera profiles for several cameras. I guess that some of the fault lies with the camera manufacturers, they probably don't give all necessary information to developers of raw processors.

Michael Reichmann and Jeff Schewe have often indicated the need for standardized raw file formats. That would make life much easier for everyone, camera vendors, developers of raw converters and also photographers.

Sorry for the long explanation, but I'm interested in both taking pictures and in the science that allows us to make pictures.

Best regards
Erik

Of course, one could calibrate any camera to a color checker in Lightroom or Camera RAW, but... then I guess the camera would adhere to Color Checker standard, which we may or may not find as pleasing. For my current Leaf back I would not dream of doing so because I find the colors very pleasing at my defaults in Capture One Pro. Colors are critical for an image, and it would seem that ten years into making of dslrs such D800 reddish hues or Canon greens should be things from past?


Best regards,
Anders

[Hening Bettermann]

you might want to consider RPP ( www.raw-photo-processor.com/ )

Hi deejjjaaaa,

thanks for the reminder of RPP, which looks really flexible. I have considered, but not tried it. I have so far stuck with Raw Developer due to its deconvolution sharpening. First if that could be done elsewhere in the pipeline would I seriously try out other converters. ACR has that now, but last time I compared (CS5), I found Raw Developer superior.
Good light!

[deejjjaaaa]

Hi deejjjaaaa,

thanks for the reminder of RPP, which looks really flexible. I have considered, but not tried it.

try - you might like what you see, subjectively of course

I have so far stuck with Raw Developer due to its deconvolution sharpening. First if that could be done elsewhere in the pipeline would I seriously try out other converters. ACR has that now, but last time I compared (CS5), I found Raw Developer superior.
Good light!

deconvolution sharpening anyways is done there (in Raw Developer) post conversion... I am yet to hear about raw converter that does sharpening somehow before demosaick/color transforms (unless it is a feature of demosaick algorithm itself somehow)... photoline is supposed to support .8bf plugins ? focus magic then might work for example

[deejjjaaaa]

Another post indicated that Capture One is quite sloppy with camera profiles, it seems that they just copied and renamed camera profiles for several cameras.

they corrected, at least further updates after that issue was raised in their U2U forum were shipping with different profiles... but that happened... I 'd assume that also happened because cameras in question were a 3rd tier (after P1 backs - 1 and after top level C&N - 2)... but who knows

[deejjjaaaa]

Michael Reichmann and Jeff Schewe have often indicated the need for standardized raw file formats. That would make life much easier for everyone, camera vendors, developers of raw converters and also photographers.

standard raw format does not guarantee a good profile there embedded, does it ?

[Anders_HK]

Hi Anders,

The idea with the color checker is to reproduce colors correctly. So, there is no color checker standard. The color checker is simply a card with 16 very well defined colors. The colors are chosen to be somewhat relevant photographically but also to have good metameric properties, what I mean be relatively consistent under different types of illumination.

So you take a picture of the color checker under two illuminations, daylight and incandescent (halogen for instance) and generate a profile that reproduces the colors as well as possible. So, in the ideal case, the color checker would reproduce exactly. Such a profile would be correct, but not necessarily pleasant. What I have seen generally that raw converters exaggerate saturation a bit.

Lars Kjellberg has produced another standard card called QPCard with accompanying software. The QPCard has a different set of color patches that may possibly more suited for portraiture.

Anyway, what I would suggest is to use either Color Checker or QPCard to generate a calibrated color and tweak the colors for pleasantness.

Another post indicated that Capture One is quite sloppy with camera profiles, it seems that they just copied and renamed camera profiles for several cameras. I guess that some of the fault lies with the camera manufacturers, they probably don't give all necessary information to developers of raw processors.

Michael Reichmann and Jeff Schewe have often indicated the need for standardized raw file formats. That would make life much easier for everyone, camera vendors, developers of raw converters and also photographers.

Sorry for the long explanation, but I'm interested in both taking pictures and in the science that allows us to make pictures.

Best regards
Erik

Erik,

I know and have color checker!!

I am only interested in one color/profile: the "neutral" or "natural" one - as accurate as possible within the technical limits of camera, screen, print, software. To this end, I have made an ICC profile for my camera using the ColorChecker and Argyll. Then I process in Raw Developer with this profile as input and output profile, everything else zeroed. Later, in my pixel editor (PhotoLine) I "normalize" to Lab. (Convert to ProPhoto for print only; to sRGB for web). As far as I can see, this is the most "accurate" color I can muster. Since I shoot landscape in daylight exclusively, I feel I can get away with just one profile.

I find that the white balance is a parameter of gross impact, but that is a topic of its own.

In my next go at improving my workflow, I look forward to  trying the QP target and software:

QPcard 202  http://www.qpcard.com/en_b2c/color-reference-cards.html

QPcalibration ICC  http://www.qpcard.com/en_b2c/applications/qpcalibration-icc-profile.html

I also look forward to explore CIECAM02.

Above is process I do not need to go through with Leaf files for Capture One Pro. Colors at defaults are already at a good departure point for my 80MP as they were for my older 28MP. Additionally there are Leaf image setting presets that can be chosen in Capture One that modifies parameters of that good departure point.

My points made; a camera should provide colors well calibrated and profiled from factory, and RAW conversion should give good departure point at defaults, with options. Good departure point should be correct colors but at same time pleasing. Thereupon makes simple to make our own further adjustments to pleasing taste for an image.

I think above is not too much to ask... Or a camera is still behind in colors compared to film.

Best regards,
Anders

[ErikKaffehr]

Hi,

I don't think it is much about camera calibration. To my best knowledge cameras are pretty constant.

It seems that several posters found issues with different raw processors. It has been also said that Nikon NX has good colors although being a crappy converter. So I don't think this is camera calibration issue, but probably more related to raw converters not supporting the camera in a proper way. That may depend on Nikon (and other vendors) not giving proper data to raw file vendors. Data that would be most helpful is probably spectral response data or color conversion matrices.

For some vendors having first class raw developers, like Phase One, keep info close to chest my be a competitive advantage. For customers having another raw conversion pipeline it is simply a disadvantage. It seems that Capture One has a history with improper profiles for cameras of lesser market share.

I'd also say that it is a great advantage if a user can stay with a single raw converter. Raw converters are complex products and it is better to have one that you know in and out.

By the way, there is a contradiction between correct colors and pleasing colors. You can have either but not both. Correct colors is what you would use for reproduction. Now, correct color is not exactly easy to achieve.

So my suggestion is that the best way may be to establish a baseline that gives correct colors and then apply a set of tweaks that gives pleasantness in your eyes. Adding pleasantness mostly includes increasing midtone gradient and adding some saturation or vibrance.

There is a company called x-equals who makes presets for different films. For my part I never liked film, really, but that was the medium we had until 2004 or so. I essentially rested my Pentax 67 around that time and used it very little since. So my history is a bit different than yours.

Best regards
Erik

Erik,
...
My points made; a camera should provide colors well calibrated and profiled from factory, and RAW conversion should give good departure point at defaults, with options. Good departure point should be correct colors but at same time pleasing. Thereupon makes simple to make our own further adjustments to pleasing taste for an image.

I think above is not too much to ask... Or a camera is still behind in colors compared to film.
...
Best regards,
Anders

[Rhossydd]

I don't think it is much about camera calibration. To my best knowledge cameras are pretty constant.

I think it's worth pointing out that we're not just talking about the camera sensor here. It's the whole optical chain, sensor > lens > filter.
When one starts to multiply up manufacturing tolerances, there could be a difference between different combinations that some people would find worth building a profile for.

[ErikKaffehr]

Hi,

That is a good point.

Best regards
Erik

I think it's worth pointing out that we're not just talking about the camera sensor here. It's the whole optical chain, sensor > lens > filter.
When one starts to multiply up manufacturing tolerances, there could be a difference between different combinations that some people would find worth building a profile for.

[Tim Lookingbill]

Achieving good color? Now there's a standard that needs to be established. Forget Raw converter camera profiles.

So what is good color? No one in this thread has demonstrated or explained what that parameter really looks like.

A color checker target is not a standard for "good color". It's a measurable target that has Lab numbers associated with it so you can compare the numbers in front of the camera to what the digital processing pipeline renders on a calibrated display. It's not a guarantee for getting accurate and/or "good color" for every scene being shot under various lighting conditions. It's a ballpark starting point. Is that good color?

Examining the incamera processing of jpegs from both a Fuji F10 P&S and Pentax K100D DSLR clearly show they tweak the color tables on output to render sky blue with either a periwinkle to cerulean to cyan hue irregardless of WB appearance. Even green foliage will have either a heavy yellowish to orangish or bluish hue. The images viewed on their own look gorgeous and somewhat what I remember how the scene appeared but with something extra.

That "something extra" is as far away from accurate as you can get after going back and examining the original landscape scene. The same thing happens with Raw except the starting point lacks that "something extra" which is really hue tweaks to cool against warm hues that imbues a harmonious color design much like the renaissance painters did.

Is that good color?

[Alan Goldhammer]

So what is good color? No one in this thread has demonstrated or explained what that parameter really looks like.

Good color is in the eye of the beholder!  Seriously, we don't take pictures of Color Checkers for display; it's a tool like any other and given the entire complexity of the image capture to print chain, aren't we trying to make things too complex?  Sure, lots of things can be tweaked and they are to varying degrees.  Ultimately it's the print that matters (and the colors within unless one is a dedicated B/W printer in which case there are other issues to address.

[Hening Bettermann]

Achieving good color?
Forget Raw converter camera profiles.

?? Then how do you output? Or do you just mean *canned* profiles?

So what is good color? No one in this thread has demonstrated or explained what that parameter really looks like.

I'll try then.

A (bottom) is a screen shot of an image with bad color. Raw with the "little extra": rosified, oversteepened, oversaturated. Raw Developer default.
B (top) is my take on "good" (at least: better) color. This satisfied my then fresh memory much more than A. My ColorChecker-based profile. WB as shot, Tint set to zero.
(Don't mind difference in crop and sharpness. A is just 1 out of 5 focus slices, B is my final image).
Good light!