Never saw the binary tree approach. And this article, being written in summer of 22, missed out on the 2D wavelet approach published later on that year.
https://cgenglab.github.io/en/publication/sigga22_wmatrix_me....
OpenCV implementations of this stuff will be complex with lots of architecture specific optimizations. They will also have tons of branches for different image formats — you can keep things very simple by just using single channel uncompressed grayscale. The Huang histogram and Perrault/Hebert constant time papers are very approachable. They have pseudocode and visualizations. They’re not that tricky. You can sit down in a day and implement them without much headache.