Question

I am using numpy is Python. I have an image loaded up into numpy 2-dimensional array as:

[
  [...], # row0
  [...], # row1
  [...], # row2
  ...
]

I need to get all the index positions of all pixels where either (only one of the following) north, south, east, or west adjacent pixels are of certain value. In my case if any of the 4 adjacent pixels is 0.

Answer 1

If a is your original array, define a bunch of slices:

from scipy import *

a = ones((12,22))
a[5,10] = a[5,12] = 0

a_ = a[1:-1, 1:-1]
aE = a[1:-1, 0:-2]
aW = a[1:-1,   2:]
aN = a[0:-2, 1:-1]
aS = a[  2:, 1:-1]

a4 = dstack([aE,aW,aN,aS])
num_adjacent_zeros = sum(a4 == 0, axis=2)
print num_adjacent_zeros

ys,xs = where(num_adjacent_zeros == 1)
# account for offset of a_
xs += 1 
ys += 1

print '\n hits:'
for col,row in zip(xs,ys):
  print (col,row)

The reason for taking the smaller a_ is that I don’t know what you want to do with the edge cases, where e.g. the North pixel might not exist.

I build an array of the count of adjacent zeros, and use that to get the positions which are adjacent to exactly one zero. Output:

[[0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]
 [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]
 [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]
 [0 0 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0]
 [0 0 0 0 0 0 0 0 1 0 2 0 1 0 0 0 0 0 0 0]
 [0 0 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0]
 [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]
 [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]
 [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]
 [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]]

 hits:
(10, 4)
(12, 4)
(9, 5)
(13, 5)
(10, 6)
(12, 6)