I took a black and white photo,
and replaced every pixel with a random color that has the same luminosity (with an extremely slow script that took like 4 hrs to run).
and this is the result.

It's not what expected, I expected it to look more random with all sorts of colors.
So I zoomed in on a little section.


Zoomed in. It does have sorts of colors but when zoom out, they tend to look like
Purple for dark tones and Yellowish Green for highlights.
So I thought well since it's not what expected, and since it has tendency to be purple-ish for dark tones and yellowish green for highlights anyways why not try to do a split tone with similar colors that show that tendency.

And that ends my findings on random colors, just thought i'd share that.
Thoughts?

Tin, about performances, did you try to use tiles or was it impossible?

nope didn't use tiles ...
just ran it and see what shows up as an experiment now that i know what it does with its random colors tendency i don't think i'll ever run it again.

Tin, another small (maybe also stupid) question: is your filter in .scm or .py?
Someone told me that the "random" function in .scm does not work as good as in .py

Dinasset,
I used .py.

Wallace,
Here's the image inverted... also something i didn't expect... interesting isn't it?
Darktones are now yellowish green, and highlights are now dark purple-ish.

I'd be interested in checking out the script if you have it available to share

Warning. you're entering a super slow zone. Use at your own risk

What best 2 colors to use to split tone an image?
I do not know but because of this experiment, I am biased towards purple and yellowish green now.
@Wallace, another reason to love purple (It represents the average of all dark random colors).

Not too surprised since our eyes are more sensitive to green. Notice that the dark tones are the complementary color...

IMHO that just means that your luminosity formula has some kind of bias... does it take in account the gamma?

You have way too much time on our hands, Tran. lololol

Who am I to judge though.

i don't think it does anything with gamma as i just calculate luminosity based on
(0.2126*R + 0.7152*G + 0.0722*B)

I revisited split-toning in GIMP. I was bored.
And found a tutorial and thought i do a simply .py script for it (There probably exists multitude of scripts already) but this one followed the tutorial i found that was like first hit on Google search.
More details about script here --> Split-toning in GIMP python script

I read through the tutorial and your examples on your script. Looks nice, will play with it tomorrow. Does it matter where we post our attempts?

Of course not...you post where you like.