掩码 Numpy 数组中的操作

Question

I have two masked numpy arrays. These are images.

I am trying to subtract one from another.

If I do the standard subtraction operator,

ma1 - ma2 

It will subtract them as if they were not masked (their masks are not considered). I would like them to subtract with their masks.

Does anyone know how to subtract masked numpy arrays from eachother?

Answer 1

It should work. When you operate on masked arrays, it takes the union of the masks involved in the operation. The case below shows how numpy chooses the values that will be changed when you do a subtraction between two masked arrays:

a1 = np.random.random((100,100))
a2 = np.random.random((100,100))

a1 = np.ma.array(a1, mask=a1<0.5)
a2 = np.ma.array(a2, mask=a2<0.5)

umask = np.logical_or(a1.mask, a2.mask) # <-- union of the masks

test = a1.data - a2.data
test[umask] = a1.data[umask] # <-- "canceling" the operation according to the
                             #     combined mask

np.allclose((a1-a2), test)
#True

As you see, the result is the same...