It's possible in Gmic something like:

(gimp-selection-none img)
(gimp-by-color-select layer
'(250 250 250) 36 0 TRUE FALSE 0 FALSE
)
(gimp-selection-invert img)
(gimp-selection-none img)

example:

[0]
-something_select_bycolor[1] rrr, ggg, bbb, %tolerance
-something_inverte_selection[1]
-operation[1] ....
-operation[2] ....
-something_selection_none[1]
-blend[0,1]


Thank's.

There is a command '-select_color' that can do this job :

Thank's.
I tried in reference.shtml but I missed -select_color ....

Some bugs?

Ok, found the solution.



Now how to use the mask [1] to apply a command/filter to [0] only in white [1] areas?
I search "mask" "alpha" in reference.shtml but nothing help me ...

Thank's.

The general idea to do that, is to apply the effect on the whole image, then copy only the masked part of the result to the initial image [0].
Example, with a blur, you can add this to your G'MIC command to do the trick.

thanks again for the tips.
I found this solution that seems to work

with -image command,
not exactly the same result ....

Because in that case, you don't have to normalize the mask image in [0,255].
The values can be left in [0,1].

Thanks again.
Same result, but with fewer steps needed:

I did some test,
a flou/soft focus filter with edges protection

the "mask" is --gimp_gradient_norm[0]

with RGBA and -blend alpha works as expected:


instead with -j the result is bad:


so I wrote the filter using the first method.
Any advice?
Thanks.

The method with '-j' assumes that your mask has values in [0,1].
For different values, you can expect weird pixels appearing. Have you checked this is indeed the case ?

understood, -j only to sharp masks [0,1].
For soft mask, alpha channel is the way.
thanks to the availability.

No, -j also works with smooth masks. But the mask values must be kept in the [0,1] interval.

ok this is a test


Mask255 [1] is a 0,255 16 level mask.
--normalize[1] 0,1
should reduce the interval in [0,1] range in new image [3]
but the result is bad ...

I suppose that the pixels of the mask should have a value of 0 or 1 (sharp mask),
not 0 ... 0.2 ... 0.8 ... 1 (soft).

Or something wrong?