how to convert a 1-dimensional image array to PIL image in Python

Question

My question is related to a Kaggle data science competition. I'm trying to read an image from a one-dimensional array containing 1-bit grayscale pixel information (0 to 255) for an 28x28 image. So the array is from 0 to 783 where each pixel is encoded as x = i * 28 + j.

Converted into a two-dimensional 28x28 matrix this:

000 001 002 003 ... 026 027
028 029 030 031 ... 054 055
056 057 058 059 ... 082 083
 |   |   |   |  ...  |   |
728 729 730 731 ... 754 755
756 757 758 759 ... 782 783

For reasons of image manipulation (resizing, skewing) I would like to read that array into an in-memory PIL image. I did some research on the Matplotlib image function, which I think is most promising. Another idea is the Numpy image functions.

What I'm looking for, is a code example that shows me how to load that 1-dimensional array via Numpy or Matplotlib or anything else. Or how to convert that array into a 2-dimensional image using for instance Numpy.vstack and then read it as an image.

Simple use np.reshape() with the desired shape of (28,28) with it? — Divakar
– Divakar, Commented Aug 15, 2016 at 9:51

unutbu · Accepted Answer · 2016-08-15 10:14:39Z

5

You can convert a NumPy array to PIL image using Image.fromarray:

import numpy as np
from PIL import Image 

arr = np.random.randint(255, size=(28*28))
img = Image.fromarray(arr.reshape(28,28), 'L')

L mode indicates the array values represent luminance. The result will be a gray-scale image.

edited Aug 15, 2016 at 10:14

answered Aug 15, 2016 at 10:07

unutbu

886k197 gold badges1.9k silver badges1.7k bronze badges

Sign up to request clarification or add additional context in comments.

5 Comments

Matthias Over a year ago

just found this one too, which is used for the Kaggle tutorial.

Santhosh Over a year ago

How do I do the same with color image stored as ndarray, where we have 3 channels BGR? @unutbu

unutbu Over a year ago

@Santhosh: To convert BGR --> RGB you could use img = img[..., ::-1].

Santhosh Over a year ago

Thank You @unutbu img = img[...,::-1] seems to be much faster than color conversion in OpenCV in which I'm working. Even this worked out = Image.fromarray(ima.reshape(ima.shape[0],ima.shape[1],3),"RGB")

Santhosh Over a year ago

Your posts have helped me a lot! Would love to learn much more tricks from you 😃 @unutbu

Collectives™ on Stack Overflow

how to convert a 1-dimensional image array to PIL image in Python

1 Answer 1

5 Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

5 Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related