When editing the Raw files in Darktable 4.2.2 (Linux) from an Olympus EM5-MK3 using the scene referred workflow, the default white balance is a bit odd. I’m not certain if this is an issue with the camera Raw support or some mistake on my part. From the camera support list, it seems the only thing my camera is missing is a custom white balance matrix.
You can observe this in the screenshots attached below:
Default scene referred workflow applied (1st screenshot).
If I change the color calibration illuminant to as shot in camera, the colors improve somewhat but still they feel a bit off. (2nd screenshot).
Finally if I apply a +0.1 change to input G in output B I get the best results so far. (3rd screenshot). I have tried this on many other shots and the colors come out much close to what I expect them to be.
I don’t have access to Adobe light room, but I know someone who does. When I open the raw file there, the colors match the jpeg produced by the camera better. I have also tried opening the files on a M1 mac with Darktable and I have the same results.
The out of camera jpeg also has a slight greenish tint to it, but I have some other examples where this is not present. I can provide more samples upon request.
Finally I also found this thread which could be somewhat related. I created this thread since the Camera is different and maybe, someone else ran into this problem as well.
Welcome to the forum. Try turning off the color calibration module and in the white balance module set to as shot in the camera. This gave a different color to doing as shot in the color calibration module. Most of the times it makes no difference, but with this image it did. It still may not be what you like.
Your JPg might have a blue boost… look at this setting in the meta…not sure but it might be adding blue?? Blue does look more saturated in the jpg… also there is some wb setting called keep the warm out…if that was used it might also we why the jpg is much colder looking… just some random guesses…
YOur image was iso 200 so I downloaded the iso 200 test image for your camera… I think DT handles wb and colors pretty well going by eye and using the vectorscope and the red green blue row of the color checker… it looks okay…maybe magenta could go more towards blue… but this is also less accurate than using data and EM5M3hSLI00200NR0.ORF (18.7 MB)
making a profile…
Hey!. Thanks. That was one of the first things I tried. It produced something closer, but still not close enough. When I tried this I also fiddled with the blue tint in the image to apply about the same correction.
These shots were indeed shot with Preserve Warm Colors set to off. I’ve found this recommendation on the dpreview forums. I’ll try to shoot a round with this setting set to warm.
That being said, even if the picture would be slightly more on the blue side. Why does it appear greenish to me? Do I have a bad monitor or is it just more yellow and I’m mistaking it for green?
I’ve also opened the reference Image, that one appears to be a bit on the warmer side of things. But no green cast.
I am not sure if this suggestion would work. But the color calibration module has a spot color mapping feature which may let you use the JPG to measure the color to set the raw file too. Works well in my hands going from one RAW image to another, but I have not yet tested the idea of JPG to RAW file.
I’ve been having the same issue with my Canon 5D mk. II. In fact, exactly the same shift towards yellow.
I initially worked around it by removing some yellow with the channel mixer, just like you. But since I created custom WB coefficients, I haven’t had to do that, although minor tweaks are usually still needed in CAT.
This video explains the process:
Note that I haven’t done the profiling part yet.
I would say it’s yellow rather than green. Greenish yellow, perhaps, but definitely yellow. Maybe it’s the yellow mixing with the blue that makes it look green to you. Is your monitor calibrated?
In this, I raised exposure a bit, used the soft contrast preset in tone eq, applied my yellow fix preset in a second instance of color calibration, added a bit of sharpness, and finally boosted the colors with color balance. Everything else is default. I’d say it’s very close to the OOC JPEG.
Hey @Donatzsky , thanks so much for the adjustment values and the video link. I tried your values on a couple of other pictures and it does indeed improve the color accuracy. There are still some outliers, but it is a step in the right direction.
I calibrated my display using a xrite colormunki. The device is almost 8 year old, maybe it’s time for an upgrade .
It looks like I have no choice but to invest in a one of those color checkers to make sure everything is properly set.
Keep in mind those values are fairly arbitrary. Basically I just found something that looked good enough in one photo and then saved the preset.
I don’t think age should be a problem with the Colormunki (unlike with the DataColor Spyder), but if you upgrade, get the CaliBrite (former X-Rite) Display Pro. I think it usually comes with a ColorChecker Mini, as well - at least mine did.
You can also try using DisplayCAL for calibration, if you didn’t already. It might give more accurate results. Here’s a good tutorial (ignore the warning about DisplayCAL being unmaintained):
For now you can take that test shot I ran and try darktable-chart. When you get your own target you can use that… It will let you create a preset to match or try to match your jpg color. The process is in the manual… This is to match JPG or a color chart… The CC module correction is more to give a correction towards the “correct” colors… I will try it for you when I get some time as there are a few steps…
I’ve got a color target, but I seem to be unable to get the right colors. My best guess is the lighting conditions are not ideal.
I’m also having an issue, where, if I follow the steps in the video regarding the exposure compensation correction, the values never update.
I found out about RawTherapee today, and I just tried to open the raw files there and the colors are displaying correctly. Could this just be a bug in Darktable?
EDIT: Some Pictures display better in RawTherapee, some also display a yellowish tint.
I’ve noticed that the E-M5 Mark II black level depends on ISO. Perhaps it is the same for Mark III (neither dt nor RT seem to have ISO specific black level data for this model)? Have you tried tweaking the black level already as well?
Alternatively, are we sure black level is read from raw ORF file metadata correctly if not from these tables?
I don’t think exposure updates the exposure is telling you what value to set exposure to so that the values calculated are accurate. So basically run the process and cycle through the profiles. Select the one with the lowest delta E and save that as a preset first changing the mode of cc to as shot. Then you can adjust exposure to the suggested value and apply it and see how it looks…
also you could try to run through the dark table chart function and see how that goes. If you share one of your test shots I can try it… there are a few steps. It was created more for an older workflow using a tone curve but the lut might be useful for cor corrections
Okay I will do one of the cc and do a darktable chart and share back and you can compare…at work now around mid day but I will do it a bit later when I am home… EST zone…
I know this isn’t something as ‘A blue boost’ or something. It affects which temperature the auto-whitebalance picks. But that’s it. So the image recorded is nothing special.
The setting tells the white-balance algorithm in the camera to neutralize colours or keel a warm tint if its there.
In heavy sunshine, it keeps the golden glow. Or in indoor shots by candlelight , it keeps the heavy yellow tint. Otherwise colours are neutralized.
Isn’t the sensor in the e-m5 mk3 very similar to the e-m1 mk3? I have that one and noticed nothing yet as far as green tint goes. Does it do something funky there with black levels or white levels or something ?
.
Thanks