Shifted white balance values for Fujifilm X-E3


(Martin Dinse) #1

Hi,
I recently got the Fuji X-E3 and am having problems getting the colors right in some lighting conditions. I tried different input profiles but that didn’t make major differences (started with an X-E3-specific adobe profile and tried the X-T20 profile that comes with RT since those two cameras basically share the components relevant for the raw data output as far as I know). Tried RT 5.3 on a Lubuntu machine as well as 5.0-r1-gtk2 on Debian. That didn’t make a difference either. The files aren’t completely unusable but they require some extra work with chromaticity and get increasingly more difficult with complicated lighting.

Yesterday I noticed that the white balance values (temperature and tint) were weirdly off. I have used RT with a variety of cameras over the years (including the Fuji X-E2, Canon, Olympus and Ricoh) and the white balance values stayed more or less the same for common lighting situations. Now for the X-E3 they are considerably off.
Let me give you an example. Sorry if this gets a bit lengthy but those are all the details I got while trying to figure out the issue myself.
I did a test shot with the X-E2 and the X-E3. Both used the same lens, same exposure and same white balance setting in camera (5900 Kelvin). I opened them in RT with no profile (neutral), both with 3-pass Trans-X demosaicking and 1 false color suppression step. Then I proceeded as follows:

For the X-E2 I:
·applied the .dcp that comes with RT
·picked the white balance from the center of the gray card, it changed from 5655/1.070 (camera setting) to 6329/1.038
·pushed the exposure +1.45 to get RGB 48.2 % reading from the gray card
·the resulting file is x-e2_DSCF1346.r.jpg

For the X-E3 I:
·applied the X-T20-.dcp that comes with RT (the adobe profile for X-E3 didn’t change the huge white balance shift but I’ll upload the output for comparison anyway)
·picked the white balance from the center of the gray card, it changed from 7988/0.521 (camera setting) to 8942/0.516; this is what I’m talking about
·pushed the exposure +1.47 to get RGB 48.2 % reading from the gray card
·the resulting file is x-e3_DSCF1444.r50.xt20.jpg

Just to illustrate the amount of white balance shift I applied the values of the X-E2 file which caused a shift of the RGB measurement in the middle of the gray card to R:27.1/G:54,1/B:35.7; it’s the picture with the heavy green cast: x-e3_DSCF1444.r50.xt20.wb2.jpg

You can find the files, accompanied by the raw files, OOC-JPGs and .pp3s here: https://drive.google.com/open?id=1X3GePa3ZNg8i1kW8cjHk2luokryCMmgM

The raw file from the X-E3 was uncompressed. With another shot I quickly compared compressed and uncompressed – there was no difference visible. Also I compared some files a friend took with his X-T20 on a recent hike with my own shots (from the same hike) and his files show ‘normal’ white balance values. So it isn’t the X-Trans III sensor in general.
I’d love to do proper test shots but unfortunately I don’t own a ColorChecker.

Does anyone have an idea what’s going on here?

Cheers,
Martn.


(Alberto) #3

might be due to the use of a suboptimal colour matrix for the x-e3.

related:

I’ll try to investigate a bit (when time allows)


(Alberto) #4

confirmed. Add this to your camconst.json (see the comment at the beginning of the file) to fix it:

    { // Quality B
        "make_model": [ "FUJIFILM X-E3" ],
        "dcraw_matrix": [ 11434,-4948,-1210,-3746,12042,1903,-666,1479,5235 ],
        "ranges": { "white": 16100 }
    }

@Morgan_Hardwood might be a good idea to include this in 5.4? (I got the matrix from darktable – I was too lazy to get it from the DNG converter…)


(Morgan Hardwood) #5

@agriggio I just did the same, getting the values from Adobe DNG Converter 10.0.0.827, and confirm that the values are correct - both matrix2 and temp and tint.

If anyone reading this has an X-E3 and a color target, now would be a great time to submit two shots.


(Morgan Hardwood) #6

Issue opened and closed:

@Martn thanks for reporting.


(Martin Dinse) #7

Thanks a lot guys.

I’ll ask around and try if I can find someone who I could borrow a color target from.