specifying semetrical kernel for Laplacian of Gaussian filter.

[MikeVeling]

My issue is about the scipy.ndimage.gaussian_laplace function in scipy.

I have a full detail write up on stack overflow here:
https://stackoverflow.com/questions/56463123/unexpected-results-in-scipy-ndimage-gaussian-laplace-inproper-2d-processing

I am attempting to apply this filter to a test image (https://imgur.com/nbfxWjB). This filter should act symmetricly, but it appears not to. For example, you get this result (https://imgur.com/3mhpXI3) with the code below. If you transpose the input and then transpose the output back you get this results (https://imgur.com/neZPDIM).

Reproducing code example:

from PIL import Image
from scipy.ndimage import gaussian_laplace

image=Image.open('White Spot.tif')
transformed=gaussian_laplace(image,4)

im = Image.fromarray(transformed)
im.save('transformed_python.TIF')

Scipy/Numpy/Python version information:

import sys, scipy, numpy; print(scipy.__version__, numpy.__version__, sys.version_info)
1.2.0 1.16.0 sys.version_info(major=3, minor=7, micro=3, releaselevel='final', serial=0)

If you were to apply a similar function in matlab you would end up with a symmetrical image

I = imread('White Spot.tif');

H = fspecial('log',[5 5],0.1);
LoG_img = imfilter(I,H,'replicate');
imshow(LoG_img);
imwrite(LoG_img,'transformed_matlab.TIF');

Results: https://imgur.com/uieZhvK.
I had to change sigma as if sigma is too big in the matlab version the image just ends up black.

I am not sure if this is a bug or if I just don't know how to specify the kernel in scipy.ndimage objects.

This issue might be related to a separate issue I saw from 2016 (#5543).

Thank you for your help!