Calculate mask from comparing values of two Numpy arrays that are different shapes

Question

Given these two arrays:

a = np.array(
    [
        [
            [1, 102, 103, 255],
            [201, 2, 202, 255],
            [201, 202, 202, 255]
        ],
        [
            [11, 120, 0, 255],
            [0, 0, 0, 255],
            [1, 22, 142, 255]
        ],
    ])

b = np.array(
    [
        [
            [1, 102, 103, 255],
            [201, 2, 202, 255]
        ],
        [
            [11, 120, 0, 255],
            [221, 222, 13, 255]
        ],
        [
            [91, 52, 53, 255],
            [0, 0, 0, 255]
        ],
    ])

a.shape # => (2, 3, 4)
b.shape # => (3, 3, 4)

I want to overlay a and b at 0, 0 and output a mask that represents when a values equal b values. The values compared are the full pixel values, so in this case [1, 102, 103, 255] is a value.

An output mask like this would be great:

result = np.array([
    [
        true,
        true,
        false
    ],
    [
        true,
        false,
        false
    ],
    [
        false,
        false,
        false
    ],
])

A perfect answer in my case would be where matching values become [255, 0, 0, 255] and mismatching values become [0, 0, 0, 0]:

result = np.array([
    [
        [255, 0, 0, 255],
        [255, 0, 0, 255],
        [0, 0, 0, 0]
    ],
    [
        [255, 0, 0, 255],
        [0, 0, 0, 0],
        [0, 0, 0, 0]
    ],
    [
        [0, 0, 0, 0],
        [0, 0, 0, 0],
        [0, 0, 0, 0]
    ],
])

result.shape # => (3, 3, 4)

It looks like this:

[![diff between a and b][1]][1]

Answer 1

Here is one possibility using slicing.

outer = np.maximum(a.shape, b.shape)
inner = *map(slice, np.minimum(a.shape, b.shape)),
out = np.zeros(outer, np.result_type(a, b))
out[inner][(a[inner]==b[inner]).all(2)] = 255,0,0,255

out
# array([[[255,   0,   0, 255],
#         [255,   0,   0, 255],
#         [  0,   0,   0,   0]],
#
#        [[255,   0,   0, 255],
#         [  0,   0,   0,   0],
#         [  0,   0,   0,   0]],
#
#        [[  0,   0,   0,   0],
#         [  0,   0,   0,   0],
#         [  0,   0,   0,   0]]])