07-17-2018, 09:25 PM
Now that I see the actual problem.... However, "this process isn't the optimal" is an assertion that requires proof. It could well be the optimal, if not for the colors, then for the processing time. A way to check this is to bluntly try all combinations of eight DMC colors, compute a "distance" of the results to the initial image, and see which one give you the smallest distance(*). If you find a result which is noticeably better than the indexing then you can invest time trying to find a better algorithm. But optimizing colors when the color balance of the initial image is possibly already overloading the WTF meter and the threads will quickly change color due to UV light (or dirt) could be seen by some as being futile.
(*) But be ready to acquire some Bitcoin-mining gear to speed this up, and you may have to come with the proper "distance" function.
(*) But be ready to acquire some Bitcoin-mining gear to speed this up, and you may have to come with the proper "distance" function.