Replicating behavior of scipy.ndimate.gaussian_filter

@GunnarFarneback Thank you very much! Playing around with padding and cropping, I found that what Scipy calls “reflect” padding is what ImageFiltering calls “symmetric”. Thus, setting this and the filter length to 9, I get the same result.

Cheers!

julia> py_filt = pyimport("scipy.ndimage").gaussian_filter(a, (1,0))
5×5 Array{Float64,2}:
 1.42704  6.42704  11.427   16.427   21.427 
 2.06782  7.06782  12.0678  17.0678  22.0678
 3.0      8.0      13.0     18.0     23.0   
 3.93218  8.93218  13.9322  18.9322  23.9322
 4.57296  9.57296  14.573   19.573   24.573 

julia> jl_filt = imfilter(a, Kernel.gaussian((1,0), (9,1)), "symmetric")
5×5 Array{Float64,2}:
 1.42704  6.42704  11.427   16.427   21.427 
 2.06782  7.06782  12.0678  17.0678  22.0678
 3.0      8.0      13.0     18.0     23.0   
 3.93218  8.93218  13.9322  18.9322  23.9322
 4.57296  9.57296  14.573   19.573   24.573 

1 Like