Hi,
I’m using Nikon D810 and tried using Darktable instead of Nikon Studio. The pictures shown in Nikon Studio present correct histogram with tiny indication of under exposure. In Darktable I do have to increase exposure to a minimum of 2.5 stops. This is creating troubles for me with the rest of the treatment.
What did I miss?
I’m using Darktable 3.8 with scene-referred and modern chromatic adaptation. I’m also using Windows 10.
Regards.
Pascal
Did you try to use Color balance RGB? Try with the preset in that module.
Hi @Pascal and welcome to the forum!
Nikon Picture Studio applies a bunch of processing by default that you can not alter. I’d guess it tries to match the in camera jpeg as a starting point. Dark table does not do that, and by default applies a pretty minimal processing to your image and expects you to do the rest. Yes! The default look of your raw file is lacking contrast, brightness, and color saturation. 2.5 stops of exposure usually means you could have collected a bit more light. I shoot a D850 and usually need to add about half a stop to a stop to get the over all brightness correct.
This is to be expected with DT and scene-referred editing…lots of good comments above…if you want to see something more similar to Nikon use the basecurve and legacy wb . This is the display referred workflow in DT…it should look more like your Nikon software. If you do want to start to do scene referred edits you will have to get used to building up the image from a more neutral starting point…often after a little experimentation you will arrive at some presets that will give you the look you want…
Still, 2.5 EV exposure correction is a lot. Seeing one of the images involved would help a lot in understanding what’s going on. E.g. a fairly dark scene with a few very bright areas close to image center could fool the camera metering…
Hi Aurélien, I already looked to your posted video. I didn’t try to get the look of jpeg produced by the camera: i didn’t shoot in jpeg, only raw. I was mainly concern by the 2.5 stops mandatory before any other processing. Perhaps was it due by wrong usage of the camera? I don’t know.
I’ll post 1 image which could be helpful.
Here is one the photo I shot during my son’s wedding.
_DSC0973.NEF (41.0 MB) directly from the camera without any modification.
I must say that I wasn’t in very good shape during the wedding due to previous operation just before it (but this wasn’t a good reason for bad work).
_DSC0973.NEF.xmp (11.2 KB)
I think the image is underexposed. Maybe the camera metered to the candle light. But it is not an issue. I did a 2min edit and used 1.8 stop EV just to show you what is possible.
If JPEG looks ok and RAW looks dark, what happens is pretty clear: the camera underexposed the shot and applied a curve to brighten back and compensate. That’s a typical trick to save highlights and avoid clipping.
So the +2.5 EV boost is nothing suprising. Whatever gives the expected result… Just put that in a preset and done.
I expect the meter was influenced by the bright candle.
A technique that may be useful is to develop the raw to a simple sRGB, and find the CLUT (Colour Look-Up Table) that would best transform that sRGB to the embedded JPEG. To find the CLUT, I use a “histogram-matching” method.
The x-axis is the input (the sRGB from the raw); the y-axis is the output (the embedded JPEG). R,G and B are shown as separate lines. We can see the overall transformation resembles a power curve. But is also raises black to a dark gray, and for highlights it raises blue and green more than red, changing those hues.
This raises 40% input to about 77% output, a factor of 1.925 in sRGB space, which is 13.3% to 55.4% in linear RGB space, which is just over 2 stops.
I’m not suggesting that this is the only tool to develop the raw to look like the JPEG, or that looking like the JPEG is a desirable goal.
But this does give some insight into a possible transformation.
That a really nice break down…may I ask what you used to grab and plot that data??
Windows commands:
%DCRAW% -v -w -T -6 -O x0.tiff _DSC0973.NEF
exiftool -jpgfromraw -b _DSC0973.NEF >x.jpg
magick x0.tiff x0.jpg
set mhDEBUG=1
call %PICTBAT%matchHisto x0.jpg x.jpg
call %PICTBAT%graphLineCol x0_mhcl.miff
%DCRAW% is my slightly hacked version of dcraw, documented at dcraw and WB.
x0.tiff is the sRGB version of the raw, and x.jpg is the embedded JPEG. Then “magick” makes an 8-bit JPEG version of x0.jpg, for matchHisto to do the real work.
Process modules: matchHisto.bat is a Windows BAT script that uses C process modules in ImageMagick to match histograms, and make the CLUT that approximately transforms the first image to the second. It writes the CLUT to x0_mhcl.miff, and Clut cookbook: graphLineCol.bat plots the channels as a graph, with the output colours shown across the bottom, perhaps confusingly.
Thanks for sharing…I appreciate you taking the time…
Perhaps I should add: I ignored the fact that the embedded JPEG is slightly cropped, and clearly denoised, and possibly sharpened. We can compensate for denoising by blurring the inputs to matchHisto, and this makes no significant difference to the CLUT.
Thanks Aurélien. My guess is: the histogram plot in Nikon Studio is not the RAW histogram but the JPEG one. Am I correct? If this assumption is correct then increasing the exposure by some stops won’t affect the following process and I can go further development in Darktable.
I’ll follow your recommendation about scene referred better than display referred. I’ll switch to jpeg or tiff only at the end of the process just when exporting the image.
I also thank the other advices. I look at them.
Currently I’ll continue development using preset for increasing global exposure.
Thank you all for using time for helping me.
Maybe you have the D-lighting activated in your Nikon camera?
Leanndruss,
you’re right. It was activated but with option “small”. I checked the photo using Nikon Studio and deactivating D-Light. The result is a little bit darker. Anyway I’ll continue using Darktable with preset (as recommended by Aurélien) of 1.5 which prevent any clipping of the highlights.
Thanks again all for your advices.
Pascal
That’s usually the case. Iirc, darktable also bases the histogram of undeveloped raws on the embedded jpeg, unless you set it to never use that jpeg (but check in the manual for pros and cons).
Keep in mind that, depending on your subjects, you could apply part or all of that correction in camera. Which has some advantages wrt shadow noise. If you do that, you’ll probably get over-exposed jpegs, though.
Thanks rvietor.
I checked the “used raw file instead of embedded jpeg…” always. This is the main histogram difference between Nikon tool and Darktable. My error came that I thought that Nikon used RAW information and not the jpeg one.
I also notice another big difference in the display environment. For DT I use “Darktable-elegent grey” in Nikon environment there is no choice: background is pure black. This makes the photo with nicer appearance on the display than with DT. Then the print step is disappointing. I definitively will get rid of Nikon Studio in the future.