first attempts with gmic-cli, garbage output?

Software G'MIC
1

looking for some more batch power, I took another look at gmic.

Got the gimp-plugin and cli-version available. I tried the latest 2.9.4 and the current ‘prerelease’ version (Windows btw).

In gimp I have one of the filters dialed in how I like. From the ‘verbose - layer name’ option, I see that what I applied is the command iain_nr_2019 1,0,0,0,0.5,1,0,8.55,3,0,3,1,0.906,4,0.

I know try to do that from the command-line.

As a test, I try the gmic -input test.tif -glow 10% -display it seems to be already.

If I try gmic -input test.tif -iain_nr_2019 1,0,0,0,0.5,1,0,8.55,3,0,3,1,0.906,4,0 -display I get some sort of garbage output:

image

Zooming in shows blocking that’s not in the source image (uncompressed 16bit tiff):
image .

Am I doing something wrong?

2

You probably need to divide by 257 first (I assume it’s a 16-bit tiff with max 65535). Also you no longer require hyphen before each command. gmic input test.tif div 257 iain_nr... and so on.

3

yeah just figured it out myself (was trying for quite some time though :frowning:).

Reading that it outputs 32bit floating-point tiff by default, it registered that what I was looking as looks like overflow, and maybe the filters want floating-point data between 0.0 and 1.0.

Why would I need to divide by 257, to get in a 0-255 range? That can’t be the default I assume.

I tried now with gmic input test.tif div 65535 ian_nr_2019 ..... mul 65535 output test2.tif,ushort and I seem to get a good output file. Is that just look?

I would understand ‘assuming floating point range’, or ‘just work with whatever the input file is’, but to assume ‘0.0 - 255.0’ seems random (and hard for filter writers :P).

Reminds me of libvips where you need to convert and cast the values to whatever is needed.

4

Yes, range handling isn’t always easy or obvious because g’mic “core” doesn’t really care. It’s left to each filter writer to decide based on needs - the expected inputs, image editor and so on. The GUI tends to hide a lot of that with either cut or normalize, so you’ll have to consider ranges more carefully in cli.

There’s a fairly steep learning curve, we’ve all been there but it’s worth it :slight_smile:

5

Iain doesn’t appear to manipulate the range; therefore, it is something that you would have to do yourself.