tasks
Karine Delvare has made a desaturate plug-in for GIMP that works just like the photoshop plug-in does.
gimp-1.0 and maybe gimp-1.2 had a desaturate that made a soft image (not unlike color processing of black and white film). I began to not use that plug-in instead using the decompose plug-in, set on HSV and then getting the Value layer. Maybe after a little levels adjustment, what had been originally a color image for me has now become digitally more similar to those old black and white prints my dad used to make.
Here, I look what the plug-in makes and see if I can read the words and reproduce it like I always suggested to others would work. I was not so successful.
On the left is the Desaturation Plug-in desaturating on lightness. My theory was that the L layer from LAB Decompose would be the same thing. Shown here on the right, it is clearly not the same.
Using the levels color tool, I managed to make the image look almost the same.
When compared on a more detailed image, my theory was (almost) completely wrong.
Once again I had some success with the levels tool.
From left to right: the original, desaturate plug-in set on luminosity, decompose layer from YCbCr R709.
From left to right: the original, desaturate plug-in set on luminosity, decompose layer from YCbCr R709. When used with a more complex image, the decompose image seemed to be somewhat soft when compared to the desaturation plugin.
From left to right: the original, desaturate plug-in set on luminosity, decompose layer from YCbCr R709 with Colors/Auto/Whitepoint filtering.
I have no idea what desaturating on luminosity really means. I do have the opinion that the whitepoint adjusted decompose luma desaturation seems to be more luminous to me. See the section of the curb below the car? I think that the glow is brighter on the hood as well.
On the left is the Desaturation Plug-in desaturating on lightness. My theory was that the L layer from HSV Decompose would be the same thing. Shown here on the right, it is clearly not the same.
Using the levels color tool, I managed to make the image look almost the same.
When compared on a more detailed image, my theory was (almost) completely wrong.
This time, working with the levels tool was not so successful. The clouds and the sky are still missing, for one example.
There is probably some paper some where which proves that these color isolating algorithms produce visually the kind of grayscale that the human eye tends to favor. After discussing this kind of mathematics with Gerald Friedman about how SIOX works, I decided to not bother finding the paper.