17

I want to convert an image to 2D array with 5 columns where each row is of the form [r, g, b, x, y]. x, y is the position of the pixel and r,g,b are the pixel values. (I will be using this array as input to a machine learning model). Is there a more efficient implementation than this in python?

import Image
import numpy as np

im = Image.open("farm.jpg")
col,row =  im.size
data = np.zeros((row*col, 5))
pixels = im.load()
for i in range(row):
    for j in range(col):
        r,g,b =  pixels[i,j]
        data[i*col + j,:] = r,g,b,i,j
Improve this question

4 Answers 4

Reset to default
20

I had to write this recently and ended up with

indices = np.dstack(np.indices(im.shape[:2]))
data = np.concatenate((im, indices), axis=-1)

Where im is a numpy array. You are probably better off reading the images straight into numpy arrays with

from scipy.misc import imread
im = imread("farm.jpg")

Or, better still if you have Scikit Image installed

from skimage.io import imread
im = imread("farm.jpg")
Improve this answer
Sign up to request clarification or add additional context in comments.

1 Comment

scipy.misc imread has been deprecated, kindly use imageio.imread()
7

I am not sure if this is the very efficient. But here you go, say arr = np.array(im); then you can do something like this.

>>> arr = np.arange(150).reshape(5, 10, 3)
>>> x, y, z = arr.shape
>>> indices = np.vstack(np.unravel_index(np.arange(x*y), (y, x))).T
#or indices = np.hstack((np.repeat(np.arange(y), x)[:,np.newaxis], np.tile(np.arange(x), y)[:,np.newaxis]))
>>> np.hstack((arr.reshape(x*y, z), indices))
array([[  0,   1,   2,   0,   0],
       [  3,   4,   5,   0,   1],
       [  6,   7,   8,   0,   2],
       [  9,  10,  11,   0,   3],
       [ 12,  13,  14,   0,   4],
       [ 15,  16,  17,   1,   0],
       [ 18,  19,  20,   1,   1],
       [ 21,  22,  23,   1,   2],
       [ 24,  25,  26,   1,   3],
       [ 27,  28,  29,   1,   4],
       [ 30,  31,  32,   2,   0],
       [ 33,  34,  35,   2,   1],
       [ 36,  37,  38,   2,   2],
       ...
       [129, 130, 131,   8,   3],
       [132, 133, 134,   8,   4],
       [135, 136, 137,   9,   0],
       [138, 139, 140,   9,   1],
       [141, 142, 143,   9,   2],
       [144, 145, 146,   9,   3],
       [147, 148, 149,   9,   4]])
Improve this answer

1 Comment

One can use numpay.asarray(im) or numpy.array(im) to convert image to numpy array.
3

I used "+" to combine two tuple, and use .append() to make "data" list.No need to use Numpy here.

row,col = im.size
data=[] #r,g,b,i,j
pixels=im.load()
for i in range(row):
  for j in range(col):
    data.append(pixels[i,j]+(i,j))
Improve this answer

Comments

1

steps are :

  1. convert images to grayscale (opencv)

  2. convert grayscale to binary image (opencv)

  3. convert to binary 2D matrix (scipy , pillow, numpy)

from scipy.ndimage import zoom
from PIL import Image
import numpy as np

srcImage = Image.open("image_in_binary_color.jpg")
grayImage = srcImage.convert('L')
array = np.array(grayImage)
array = zoom(array, 310/174)

np.savetxt("binarized.txt", array<128, fmt="%d")
print("\n\n Output Stored to binarized.txt.......#")
  1. store it in a file named binarized.txt

This is how i did it : https://github.com/jithi22/Imagery.git

Improve this answer

Comments

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.