I use a gaussian filter to preprocess microscope images, something like
img=filters.gaussian(imgrescaled,sigma=gaussradius)
but recently I needed to perform an additional filtering on a selected region (circular, x0,y0, radius r), where I'd like to smooth out a defect.
Unfortunately I discovered from https://scikit-image.org/docs/dev/api/skimage.filters.html#skimage.filters.gaussian that while many filters provide you with an optional mask argument, this is not true for the gaussian filter.
There is some alternative?
You could copy the image, preform the additional filtering on the copy, crop the copy image to your desired region and replace the region in your old image with the cropped region. What data structure are you using for your images?