Optimize calculation of statistics over rows of a matrix in Python + Numba
I have a code which basically calculates the mean of the channels of each channel of an image.
Yet the mean is a robust mean calculated between quantiles and iqnores zero pixels (all channels value is 0):