THX for the responses
I hadn't mentioned, but I'm using "-2014/08/20 : version 1.6.0.0 (current stable)" per the release notes under GMIC about
Played some more with this today and find myself in engineering hell (I am now unable to replicate the top half to blue channel behavior) {Changing the preview type to "Duplicate Horizontal" (which I remember using in several of the early attempts) does not have any effect}
Wallace wrote:
It could be that this particular jpg file is corrupt in some way.
Have you tried to convert the image to say a png file or a gif file and then trying to use that G'Mic filter on it?
I'd previously tried both
— exporting the file from GIMP as a jpg (just open the file then export it under a different name)
— exporting the file after duplicating the layer and setting the duplicate to Overlay100%
I've tried your suggestion and converted the original to a png (both with GIMP and with IrfanViewer); both png files turn completely black (#000000) after applying the filter.
Iain wrote:
Sounds like a problem with the parallel processing. Does it still happen if you set parallel processing to one thread?
Interesting idea, but unfortunately get full black regardless of the parallel processing setting. (Used auto & one thread on each test image: the original, & each png created per Wallace's suggestion. )
Ronounours wrote:
Is this possible to get a test case (image + parameters used) to be able to reproduce the problem ?
Thanks !
Base image cannot be provided
"Your post looks too spamy for a new user, please remove off-site URLs." All one (1) of them

though if you were to check imagebam image b3361b387760459 you could find it
Unfortunately this is a reupload/rehosting of my copy -- I can't find anything to tell me where i dowloaded the image from…
Parameters are default for the filter (other than trials with Iain's suggested one thread change).
[Sigma 1.00
Iterations 10
"Cut" checked
Parallel processing Auto (or One thread - see above)
Spatial overlap 24
Preview type Full]
I'd done some more or less random changing when I first hit this problem, but neglected to take notes
I had made some trials using either the top half or either of the top quarters (1/2 vert x 1/2 hrz) copied and pasted as a new image with the thinking the issue might be memory, but these all went to black as well :scratch: