Interesting, thanks for the details!

On 02/18/2018 04:15 AM, johannes hanika wrote:
hi robert,

glad to hear it worked in the end :)

the base- vs tonecurve thing is a philosophical question to some
extent. the main difference is the position in the pipeline: a base
curve makes everything non-linear at a quite early point in the
pipeline. this can be desirable, for instance to compress hdr via
exposure fusion and then work on that with the other tools.

for linear workflows, the tonecurve may be the better choice as it
comes later in the pipeline. when working on shadows/highlights/local
contrast with the "local contrast" module for instance i get much
improved rendition with the tonecurve.

the other big difference is that tonecurve is Lab and basecurve is
camera rgb. with the "auto in RGB" colour adjustment option in the
tone curve, you have pretty much an rgb curve here, too. for the
colour lut to be somewhat portable in theory (and mostly for less
saturated colours where the matrix profiles kind of work best..), the
lut module was implemented in device-independent Lab colour space.
that way you could profile the fuji film emulation styles once and
then apply to your canon or nikon cameras. this only holds when you
profile against pfm which have the matrix profile (from the input
colour profile module) baked in, and of course only within the limits
of matrix profiles.

my guess would be that even if you use an it8 chart, the colour-lut
tone curve would be a bit more approximate than the one you get with
the dedicated curve module? just because there is a limited number of
grey patches in the chart.

hope that helps,

On Sat, Feb 17, 2018 at 10:43 PM, Robert Bieber <> 
Today I finally got some free time to try shooting some calibration photos
with my new camera, and I did two things.  Well, three, kind of.

1.First I tried to use the darktable-chart tool to build a full color
correction profile, but this didn't work and I couldn't figure out why,

2. I used the base curve tool with some test chart images and a photo of a
white wall + grid spotted light + some black posterboard and actually came
away with a pretty good base curve that got me a lot closer to what I saw in
the back of my camera (and with the manufacturer's RAW processor if I tried
that) than any of the built-in profiles had done (and rightly so, it looks a
lot different from any of them).

3. I finally realized my mistake from step 1, which is that I'd tried just
using imagick to convert a reference jpeg to pfm, when I needed to import it
to DT so I could export it in Lab colorspace.  Once that was done I managed
to generate a complete style from a color checker shot.

The generated style with its settings for the color LUT module has
drastically improved the color rendering for RAW files from this camera.
What I'm curious about is that the generated style applies a very similar
curve to the base curve that I generated in step 2, but it applies it as a
tone curve and turns the base curve off.

Does anyone know why it does this?  I thought you generally wanted to apply
a base curve that more or less matched the one the manufacturer uses, as a
starting point, but when we generate these styles we're not using one at
all.  I noticed that if I tried turning off the tone curve and applying the
base curve that I generated it seems to make the color LUT do some weird
things, so I'm guessing that has something to do with it?

darktable user mailing list to unsubscribe send a mail to

darktable user mailing list
to unsubscribe send a mail to

Reply via email to