Counting of adjacent cells in a numpy array

Question

Past midnight and maybe someone has an idea how to tackle a problem of mine. I want to count the number of adjacent cells (which means the number of array fields with other

Answer 1

I think you already got it. after dilation, the number of 1 is 19, minus 5 of the starting shape, you have 14. which is the number of zeros surrounding your shape. Your total of 24 has overlaps.

Answer 2

Use a convolution to count neighbours:

import numpy
import scipy.signal

a = numpy.zeros((6,6), dtype=numpy.int) # Example array
a[2:4, 2:4] = 1;a[2,4] = 1 # with example value structure

b = 1-a
c = scipy.signal.convolve2d(b, numpy.ones((3,3)), mode='same')

print numpy.sum(c * a)

b = 1-a allows us to count each zero while ignoring the ones.

We convolve with a 3x3 all-ones kernel, which sets each element to the sum of it and its 8 neighbouring values (other kernels are possible, such as the + kernel for only orthogonally adjacent values). With these summed values, we mask off the zeros in the original input (since we don't care about their neighbours), and sum over the whole array.