For ml data that has been clipped and masked, add in feature so the NPI interpolation doesn't extend to areas that won't be needed (eg if coherence mask goes to edge of image, no need to interpolate to there as a) that's slower and b) it can create corrections that are applied to nothing)
For ml data that has been clipped and masked, add in feature so the NPI interpolation doesn't extend to areas that won't be needed (eg if coherence mask goes to edge of image, no need to interpolate to there as a) that's slower and b) it can create corrections that are applied to nothing)