Show HN: IGN – Convert your image in any color palettes(ign.schrodinger-hat.it) |
Show HN: IGN – Convert your image in any color palettes(ign.schrodinger-hat.it) |
The algorithm you've used is a Euclidean sum of components https://github.com/Schrodinger-Hat/ImageGoNord-pip/blob/mast... You may (human perception!) find better colour accuracy in using the Compuphase algorithm https://www.compuphase.com/cmetric.htm or possibly a luminance or other cielab approach.
The implementation of the former is not much more difficult, e.g. see the C version breakout box on that link, but you should find (at least for my projects where I've used it) colour quantisation is (subjectively!) substantially better.
You're right, there is a big room for improvement there.
We choose the first one at the beginning but it might look better (or we can even let the user choose) which algorithm perform.
I'm going to work on that! Again, thanks