[tgb]

Reading how this works, I'm a bit disappointed in how the low sample images performs, particularly for the stems. It feels like one should be able to tweak this to get much better results in this situation. I'm disappointed in the amount of yellow over the stem.

Here's my first thought on what could improve that: samples are taken at edges, which is exactly where colors vary quickly. So perhaps samples should be taken in pairs, one on either side of the edge.

Fun project!

[TheLoneWolfling]

The "easy" way to do this is to run the edge detection twice. In other words, run edge detection, then run edge detection on the result of the edge detection.