To make GLCM symmetrical, the python code performed an inplace addition: `P_glcm += numpy.transpose(P_glcm, (1, 0, 2))`. However, when the matrix exceeds a certain size, this produces buggy behaviour, where for some elements, the calculated value equals `P_glcm += P_glcm + numpy.transpose(P_glcm, (1, 0, 2))`. This is caused by the fact that numpy.transpose returns a *view*, and not a copy of the array. Then, if the additions have to be processed in multiple chunks, the addition will become erroneous.

Use `numpy.copy` to get a transposed copy of the GLCM and not just a view of the GLCM.