specific application of gaussian

I’m writing an application that benefits immensely from high level gaussian filters. However, because of the computational load of that filter, I want to apply it in only a specific area of interest. Since it appears that the img.gaussian() doesn’t take a roi argument, how do I generate a mask image manually? All I’d need is a way to generate a pure black image, then write specific white rectangles in areas.
Also, am I correct in that pixels that are masked out are not blurred at all (the processor doesn’t even run the function on them), thereby cutting that compute time out?

Hi, we’ve optimized the firmware in the latest head of our repo to make that go way faster. Median filter for example was speed up 1000%.

Um, attached is a firmware binary. Give some of the filters a try. For the H7.
firmware.zip (1.16 MB)