Re: [Gimp-user] tone mapping: gimp vs gegl CLI

[Matthew Woehlke <mwoehlke floss gmail com>]

On 2018-08-07 15:47, Matthew Woehlke via gimp-user-list wrote:
> I want to apply mantiuk06 to a really big image. However, using the 
> GEGL CLI, I get very different results than when I apply the filter
> from within gimp as compared to using the GEGL CLI. See [1]: left is
> original, middle is mantiuk06 applied from within gimp, right is 
> mantiuk06 applied via GEGL CLI.
>
> What is gimp doing differently from GEGL CLI, and is there a way to get
> the gimp results with the GEGL CLI?
>
> [1] https://gitlab.gnome.org/GNOME/gegl/uploads/e1e40b302523405d10f1234eec1edbeb/roses.png

To forward Massimo's answer from
https://gitlab.gnome.org/GNOME/gegl/issues/94, it seems gimp (2.8) and
GEGL CLI disagree about whether the image is in a linear or
gamma-corrected color space, and one is adding an implicit conversion.

While this is an "improvement" for some operations (for mantiuk06, it's
hard to argue that the GEGL result isn't more "correct", although
stylistically I prefer the "wrong" result in some cases), for many
others it ranges from dubiously, to *definitely*, worse. In particular,
stress tends to produce a washed-out result. (I mention this mainly
because it is my impression that gimp 2.10 behaves the same was as GEGL
CLI, though it sounds as if there is an option to control this. PLEASE
KEEP THAT OPTION!)

At any rate, the literal answer to my question is:

  gegl <in> -o <out> \
    cast-format input-format="R'G'B' u8" output-format="RGB u8" \
    <operation> <options> \
    cast-format input-format="RGB u8" output-format="R'G'B' u8"

...which does indeed produce the same results as gimp (2.8). \o/

-- 
Matthew