Faster way to extract patches from images?

Question

I am trying to extract patches of fixed size centered at some given position (x,y). The code is given below-

for i,j in zip(indices[0],indices[1]):
    patch         


        

Answer 1

Assuming you are dealing with near-boundary indices separately, as otherwise you would have different shaped patches, let us suggest ourselves a vectorized approach making use broadcasting together with some knowledge about linear-indexing. Posted below is an implementation to go with that philosophy to give us a 3D array of such patches -

m,n = x.shape
K = int(np.floor(patch_size/2.0))
R = np.arange(-K,K+1)                  
out = np.take(x,R[:,None]*n + R + (indices[0]*n+indices[1])[:,None,None])

Let's do a sample run on a minimal input case with input image x of (8,10) and indices are such that the desired patches don't extend beyond the boundaries of the input image. Then, run the original and proposed approaches for verification. Here we go -

1] Inputs :

In [105]: # Inputs
     ...: x = np.random.randint(0,99,(8,10))
     ...: indices = np.array([[4,2,3],[6,3,7]])
     ...: 

3] Original approach with output :

In [106]: # Posted code in the question ...

In [107]: patches[0]
Out[107]: 
array([[[92, 21, 84],
        [10, 52, 36],
        [ 5, 62, 61]]])

In [108]: patches[1]
Out[108]: 
array([[[71, 76, 75],
        [80, 32, 55],
        [77, 62, 42]]])

In [109]: patches[2]
Out[109]: 
array([[[16, 88, 31],
        [21, 84, 51],
        [52, 36,  3]]])

3] Proposed approach with output :

In [110]:  # Posted code in the solution earlier ...

In [111]: out
Out[111]: 
array([[[92, 21, 84],
        [10, 52, 36],
        [ 5, 62, 61]],

       [[71, 76, 75],
        [80, 32, 55],
        [77, 62, 42]],

       [[16, 88, 31],
        [21, 84, 51],
        [52, 36,  3]]])

Answer 2

Using scikit-learn:

from sklearn.feature_extraction.image import extract_patches

all_patches = extract_patches(x, patch_size)

upper_left = indices - patch_size // 2
patches = all_patches[upper_left[0], upper_left[1]]

A similar function can be found in scikit-image: view_as_windows.