Image padding with reflect (mirror) in Python

Question

I have an image loaded in python as an np.array with shape (7, 7). I need to apply a filtering process to the image. For this I need to apply a kernel that will convolve the 2D image (like a moving window). However, to get an output filtered image with the same shape as the input I need to expand the original image (padding) and fill these pixels as a mirror before filtering.

Below I illustrate my problem with an image:

Note: Padding must be applied to all corners of the image.

How can I create this padding?

Here some example data:

img = np.array([[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, 37, 39, 40, 41, 42],
                [43, 44, 45, 46, 47, 48, 49]], dtype=np.uint8)

plt.imshow(img)

Asked By: sermomon

||

Source

Answer 1

We may use np.pad with mode = 'reflect':

img = np.pad(img, ((2, 2), (2, 2)), 'reflect')

Example:

import numpy as np

img = np.array([[ 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, 37, 39, 40, 41, 42],
                [43, 44, 45, 46, 47, 48, 49]], dtype=np.uint8)

pimg = np.pad(img, ((2, 2), (2, 2)), 'reflect')

Value of pimg:

array([[17, 16, 15, 16, 17, 18, 19, 20, 21, 20, 19],
       [10,  9,  8,  9, 10, 11, 12, 13, 14, 13, 12],
       [ 3,  2,  1,  2,  3,  4,  5,  6,  7,  6,  5],
       [10,  9,  8,  9, 10, 11, 12, 13, 14, 13, 12],
       [17, 16, 15, 16, 17, 18, 19, 20, 21, 20, 19],
       [24, 23, 22, 23, 24, 25, 26, 27, 28, 27, 26],
       [31, 30, 29, 30, 31, 32, 33, 34, 35, 34, 33],
       [37, 37, 36, 37, 37, 39, 40, 41, 42, 41, 40],
       [45, 44, 43, 44, 45, 46, 47, 48, 49, 48, 47],
       [37, 37, 36, 37, 37, 39, 40, 41, 42, 41, 40],
       [31, 30, 29, 30, 31, 32, 33, 34, 35, 34, 33]], dtype=uint8)

Answered By: Rotem

Image padding with reflect (mirror) in Python

Question:

Answers: