crop center portion of a numpy image

Question

Let's say I have a numpy image of some width x and height y. I have to crop the center portion of the image to width cropx and height cropy. Let's assume that cropx and cropy are positive non zero integers and less than the respective image size. What's the best way to apply the slicing for the output image?

score 69 · Accepted Answer · answered Sep 08 '16 at 04:12

Something along these lines -

def crop_center(img,cropx,cropy):
    y,x = img.shape
    startx = x//2-(cropx//2)
    starty = y//2-(cropy//2)    
    return img[starty:starty+cropy,startx:startx+cropx]

Sample run -

In [45]: img
Out[45]: 
array([[88, 93, 42, 25, 36, 14, 59, 46, 77, 13, 52, 58],
       [43, 47, 40, 48, 23, 74, 12, 33, 58, 93, 87, 87],
       [54, 75, 79, 21, 15, 44, 51, 68, 28, 94, 78, 48],
       [57, 46, 14, 98, 43, 76, 86, 56, 86, 88, 96, 49],
       [52, 83, 13, 18, 40, 33, 11, 87, 38, 74, 23, 88],
       [81, 28, 86, 89, 16, 28, 66, 67, 80, 23, 95, 98],
       [46, 30, 18, 31, 73, 15, 90, 77, 71, 57, 61, 78],
       [33, 58, 20, 11, 80, 25, 96, 80, 27, 40, 66, 92],
       [13, 59, 77, 53, 91, 16, 47, 79, 33, 78, 25, 66],
       [22, 80, 40, 24, 17, 85, 20, 70, 81, 68, 50, 80]])

In [46]: crop_center(img,4,6)
Out[46]: 
array([[15, 44, 51, 68],
       [43, 76, 86, 56],
       [40, 33, 11, 87],
       [16, 28, 66, 67],
       [73, 15, 90, 77],
       [80, 25, 96, 80]])

Losses Don · Answer 2 · 2018-06-12T06:51:35.700

A more general solution based on @Divakar 's answer:

def cropND(img, bounding):
    start = tuple(map(lambda a, da: a//2-da//2, img.shape, bounding))
    end = tuple(map(operator.add, start, bounding))
    slices = tuple(map(slice, start, end))
    return img[slices]

and if we have an array a

>>> a = np.arange(100).reshape((10,10))

array([[ 0,  1,  2,  3,  4,  5,  6,  7,  8,  9],
       [10, 11, 12, 13, 14, 15, 16, 17, 18, 19],
       [20, 21, 22, 23, 24, 25, 26, 27, 28, 29],
       [30, 31, 32, 33, 34, 35, 36, 37, 38, 39],
       [40, 41, 42, 43, 44, 45, 46, 47, 48, 49],
       [50, 51, 52, 53, 54, 55, 56, 57, 58, 59],
       [60, 61, 62, 63, 64, 65, 66, 67, 68, 69],
       [70, 71, 72, 73, 74, 75, 76, 77, 78, 79],
       [80, 81, 82, 83, 84, 85, 86, 87, 88, 89],
       [90, 91, 92, 93, 94, 95, 96, 97, 98, 99]])

We can clip it with cropND(a, (5,5)), you will get:

>>> cropND(a, (5,5))

array([[33, 34, 35, 36, 37],
       [43, 44, 45, 46, 47],
       [53, 54, 55, 56, 57],
       [63, 64, 65, 66, 67],
       [73, 74, 75, 76, 77]])

It not only works with 2D image but also 3D image.

Have a nice day.

why is this not more upvoted? Images can quite often have multiple channels (3D) — simplename, Jun 07 '18 at 20:59
Hi, I have 3D image with the shape of (WxHxD) (281, 389, 104), but once I am trying to run `cropND`, I receive an error: `Traceback (most recent call last): File "", line 1, in cropND(img,(256,256)) File "", line 2, in cropND start = tuple(map(lambda a, da: a//2-da//2, img.shape, bounding)) File "", line 2, in start = tuple(map(lambda a, da: a//2-da//2, img.shape, bounding)) TypeError: unsupported operand type(s) for //: 'NoneType' and 'int'` — S.EB, Jul 13 '18 at 10:09

score 2 · Answer 3 · edited Oct 03 '18 at 17:29

2

Thanks, Divakar.

Your answer got me going the right direction. I came up with this using negative slice offsets to count 'from the end':

def cropimread(crop, xcrop, ycrop, fn):
    "Function to crop center of an image file"
    img_pre= msc.imread(fn)
    if crop:
        ysize, xsize, chan = img_pre.shape
        xoff = (xsize - xcrop) // 2
        yoff = (ysize - ycrop) // 2
        img= img_pre[yoff:-yoff,xoff:-xoff]
    else:
        img= img_pre
    return img

edited Oct 03 '18 at 17:29

synic

26,359
20
111
149

answered Sep 08 '16 at 05:24

Gert Gottschalk

1,658
3
25
37

You might want to use `//` instead of `/` to avoid making them floating pt numbers as they might not work for indexing, at least I guess with Python 3.x versions. – Divakar Sep 08 '16 at 05:29
I would remove the questions from this "answer" – simplename Jun 07 '18 at 20:58
1

This does not work for odd numbers. E.g. cropping a 101 wide image to 50 will result in a 51 wide image! – Flamefire Nov 20 '18 at 09:37

score 2 · Answer 4 · answered Jul 29 '19 at 05:44

A simple modification from @Divakar 's answer that preserves the image channel:

    def crop_center(self, img, cropx, cropy):
       _, y, x = img.shape
       startx = x // 2 - (cropx // 2)
       starty = y // 2 - (cropy // 2)
       return img[:, starty:starty + cropy, startx:startx + cropx]

score 2 · Answer 5 · answered Oct 22 '21 at 14:10

I had a problem in which I needed to crop the center from both 2D and 3D arrays. That meant unpacking a varying number of items from img.shape.

Dropping this here in case someone runs into the same problem. This modification of Divkar's solution allows cropping 2D OR 3D arrays

def crop_center(img, cropx, cropy):
    y, x, *_ = img.shape
    startx = x // 2 - (cropx // 2)
    starty = y // 2 - (cropy // 2)    
    return img[starty:starty + cropy, startx:startx + cropx, ...]

score 1 · Answer 6 · answered Nov 06 '20 at 10:59

1

Another simple modification from @Divakar's answer to preserve color channels:

def crop_center(img,cropx,cropy):
    y,x,_ = img.shape
    startx = x//2-(cropx//2)
    starty = y//2-(cropy//2)
    return img[starty:starty+cropy,startx:startx+cropx,:]

answered Nov 06 '20 at 10:59

leenremm

1,083
13
19

crop center portion of a numpy image

6 Answers6

Linked

Related