Clever Geek Handbook

Fastest way to select 7 * 7 adjacent pixels for every pixel in an image in Python

need to read the image as an array, and for each pixel select 7 * 7 adjacent pixels, then change it and put it as the first row of the workout set:

  import numpy as np
  from scipy import misc
  face1=misc.imread('face1.jpg')

face1

sizes (288, 352, 3)

, you need to find 7 * 7 adjacent pixels for each pixel, so the color 49 * 3 will then change it as an array (1,147)

and put it in an array for all pixels, I took the following approach:

X_training=np.zeros([1,147] ,dtype=np.uint8)
for i in range(3, face1.shape[0]-3):
    for j in range(3, face1.shape[1]-3):
        block=face1[i-3:i+4,j-3:j+4]
        pxl=np.reshape(block,(1,147))
        X_training=np.vstack((pxl,X_training))

final form X_training

(97572, 147)

and the last line contains all zeros:

a = len(X_training)-1
X_training = X_training[:a]

the above code works well for one image, but since Wall time: 5min 19s

I have 2000 images, so it will take quite a while for all images. I'm looking for a faster way to iterate over each pixel and accomplish the above task.

Edit: enter image description here this is what I mean by neighboring pixels, for each pixelface1[i-3 : i+4 ,j-3:j+4]

+3


source to share


3 answers


An efficient way is to use stride_tricks

to create a 2d rolling window over the image, then flatten it:

import numpy as np

face1 = np.arange(288*352*3).reshape(288, 352, 3)  # toy data

n = 7  # neighborhood size

h, w, d = face1.shape
s = face1.strides

tmp = np.lib.stride_tricks.as_strided(face1, strides=s[:2] + s,
                                      shape=(h - n + 1, w - n + 1, n, n, d))
X_training = tmp.reshape(-1, n**2 * d)
X_training = X_training[::-1]  # to get the rows into same order as in the question


tmp

is a 5D representation in an image, where it is tmp[x, y, :, :, c]

equivalent to a neighborhood face1[x:x+n, y:y+n, c]

in a color channel c

.

+4


source


Below is <1 s on my laptop:

import scipy as sp
im = sp.rand(300, 300, 3)

size = 3
ij = sp.meshgrid(range(size, im.shape[0]-size), range(size, im.shape[1]-size))
i = ij[0].T.flatten()
j = ij[1].T.flatten()

N = len(i)
L = (2*size + 1)**2
X_training = sp.empty(shape=[N, 3*L])

for pixel in range(N):
    si = (slice(i[pixel]-size, i[pixel]+size+1))
    sj = (slice(j[pixel]-size, j[pixel]+size+1))
    X_training[pixel, :] = im[si, sj, :].flatten()

X_training = X_training[-1::-1, :]


I'm always sad when I can't think of a one-line vector version, but at least it's faster for you.

+3


source


Using scikit-image:

import numpy as np
from skimage import util

image = np.random.random((288, 352, 3))
windows = util.view_as_windows(image, (7, 7, 3))

out = windows.reshape(-1, 7 * 7 * 3)
+3


source






More articles:


All Articles