inverting image in Python with OpenCV

Question:

I want to load a color image, convert it to grayscale, and then invert the data in the file.

What I need: to iterate over the array in OpenCV and change every single value with this formula (it might be wrong but it seems reasonable for me): 

img[x,y] = abs(img[x,y] - 255)

but I don’t understand why doesn’t it works:

def inverte(imagem, name):
    imagem = abs(imagem - 255)
    cv2.imwrite(name, imagem)


def inverte2(imagem, name):
    for x in np.nditer(imagem, op_flags=['readwrite']):
        x = abs(x - 255)
    cv2.imwrite(name, imagem)


if __name__ == '__main__':
    nome = str(sys.argv[1])
    image = cv2.imread(nome)
    gs_imagem = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
    inverte(gs_imagem, "invertida.png")
    inverte2(gs_imagem, "invertida2.png")

I don’t want to do an explicit loop (I am trying to be more pythonic). I can see that in one image that got a white background it turned black, but only this it doesn’t looks like the other colors are having much (if any) change.

Asked By: Mansueli

||
Source

Answers:

You almost did it. You were tricked by the fact that abs(imagem-255) will give a wrong result since your dtype is an unsigned integer. You have to do (255-imagem) in order to keep the integers unsigned:

def inverte(imagem, name):
    imagem = (255-imagem)
    cv2.imwrite(name, imagem)

You can also invert the image using the bitwise_not function of OpenCV:

imagem = cv2.bitwise_not(imagem)
Answered By: Saullo G. P. Castro

Alternatively, you could invert the image using the bitwise_not function of OpenCV:

imagem = cv2.bitwise_not(imagem)

I liked this example.

Answered By: Eric Olmon

You can use “tilde” operator to do it:

import cv2
image = cv2.imread("img.png")
image = ~image
cv2.imwrite("img_inv.png",image)

This is because the “tilde” operator (also known as unary operator) works doing a complement dependent on the type of object

for example for integers, its formula is:

x + (~x) = -1

but in this case, opencv use an “uint8 numpy array object” for its images so its range is from 0 to 255

so if we apply this operator to an “uint8 numpy array object” like this:

import numpy as np
x1 = np.array([25,255,10], np.uint8) #for example
x2 = ~x1
print (x2)

we will have as a result: 

[230 0 245]

because its formula is:

x2 = 255 – x1

and that is exactly what we want to do to solve the problem.

Answered By: Diroallu

You can also do it with numpy.

import cv2
import numpy as np

image = cv2.imread('your_image', 0)
inverted = np.invert(image)

cv2.imwrite('inverted.jpg', inverted)
Answered By: Lakshman

Why not use the first line in the question with numpy?

inverted = np.abs(image - 255)

Just as simple as that. No iteration or any other function needed. numpy does that automatically for us 🙂

Answered By: theProcrastinator

In Python/OpenCV, I think you want:

img = cv2.absDiff(img, 255)
Answered By: fmw42

Don’t use

np.invert()

to invert a binary image. it changes the value of you image from [0,1] to 255. instead using:

inverted_image= cv2.bitwise_not(binary_image)
Answered By: omid