Scale a numpy array with from -0.1 - 0.2 to 0-255

Question

I have an numpy array in python that represent an image its size is 28x28x3 while the max value of it is 0.2 and the min is -0.1. I want to scale that image between 0-255. How can I do so?

You may want to check the answers here: https://stackoverflow.com/questions/21030391/how-to-normalize-an-array-in-numpy — Leonardo Gonzalez, Apr 19 '18 at 13:38

score 24 · Accepted Answer · edited Jun 20 '20 at 13:43

24

new_arr = ((arr + 0.1) * (1/0.3) * 255).astype('uint8')

This first scales the vector to the [0, 1] range, multiplies it by 255 and then converts it to uint8, which is a common format for images (opencv uses it, for example)

In general you can use:

new_arr = ((arr - arr.min()) * (1/(arr.max() - arr.min()) * 255)).astype('uint8')

edited Jun 20 '20 at 13:43

Milan Cermak

7,476
3
44
59

answered Apr 19 '18 at 13:37

Fred

1,462
8
15

Yes indeed the max is 0.2 and not 0.3. Thanks I will check it out. – Jose Ramon Apr 19 '18 at 13:39
1

`0.3` is the right number in that code. Check again the generalized version – Fred Apr 19 '18 at 13:40
It's probably a bit better (though ugly) to use s.th. like `255.999` to have equal bin sizes. – Paul Panzer Apr 19 '18 at 13:59
Probably better to just use `np.round` – Fred Apr 19 '18 at 14:05

BHT · Answer 2 · 2018-09-23T19:24:26.347

4

You can also use uint8 datatype while storing the image from numpy array.

import numpy as np from PIL import Image img = Image.fromarray(np.uint8(tmp))

tmp is my np array of size 255*255*3.

edited Sep 23 '18 at 19:24

answered Sep 23 '18 at 19:06

BHT

41
4

Scale a numpy array with from -0.1 - 0.2 to 0-255

2 Answers2