I have a gray-scale image in Gimp and I want to process it like this:
*multiply image by 2
*subtract 1
*square the image (i.e. image*image)

In short output = (1-image*2)^2

So if have the following input image:

The expected result is:


The math is dead simple and it works if I use math nodes in Blender,
so the problem is my G'MIC implementation:


I thought maybe the problem is G'MIC processing the image as 8bit integer
instead of float, but if I try to convert it with -int82float I get an error.

Someone knows how to implement this equation in G'MIC?
Maybe you have to use int82float but you can't because Gimp doesn't support float images?
Is there a workaround to get the same result?

Even if G'MIC does all its computation using float values, it reads pixel values as they are stored in the image file, so for a .JPEG file, the pixel values will be probably in range [0-255]
(few JPEG files store data in 12bits integers, but it is rarely the case). GIMP also transfer image data to the G'MIC plug-in as 8bits integers.
Actually, you cannot assume that the values are normalized in [0-1], and you must explicitly divide the values by 255 and re-multiply them afterwards, in your processing code if you need too :



Note that you can also uses the '-fill' command, with a mathematical expression. It is probably a bit slower, but easier to read :

Thank you!

I thought G'MIC loaded and worked on images as floats, converting everything to 8bit when bringing back the results to Gimp.

I didn't know you can access the pixel value from a mathematical expression, nice tip.