Hello, ...

A part from this the algorithm could be better. From my test using stddev and sum2 yield better result.

Maybe mean filtering the parts before processing could also be a good idea? Should dim moving artefacts. -- Happy coding :) Vibe