Converting an image to grayscale using numpy

Question

I have an image represented by a numpy.array matrix nxm of triples (r,g,b) and I want to convert it into grayscale, , using my own function.

My attempts fail converting the matrix nxmx3 to a matrix of single values nxm, meaning that starting from an array [r,g,b] I get [gray, gray, gray] but I need gray.

i.e. Initial colour channel : [150 246 98]. After converting to gray : [134 134 134]. What I need : 134

How can I achieve that?

My code:

def grayConversion(image):
    height, width, channel = image.shape
    for i in range(0, height):
        for j in range(0, width):
            blueComponent = image[i][j][0]
            greenComponent = image[i][j][1]
            redComponent = image[i][j][2]
            grayValue = 0.07 * blueComponent + 0.72 * greenComponent + 0.21 * redComponent
            image[i][j] = grayValue
    cv2.imshow("GrayScale",image)
    return image

I'm not sure if I understand your quetion right. What's exactly the problem with the opencv conversion? Do you want to implement the opencv conversion yourself? — xFL
– xFL, Commented Jul 11, 2018 at 13:01
There is no any problem with opencv conversion. Yes, I wanted a 2-dimensional array of image implementing my own code. — thesamiroli
– thesamiroli, Commented Jul 11, 2018 at 13:03
thesamiroli if my answer isn't correct could you please expand on what am I missing? Thanks! — Luca Cappelletti
– Luca Cappelletti, Commented Jul 11, 2018 at 13:13

Jeru Luke · Accepted Answer · 2018-07-11 13:49:40Z

5

Here is a working code:

def grayConversion(image):
    grayValue = 0.07 * image[:,:,2] + 0.72 * image[:,:,1] + 0.21 * image[:,:,0]
    gray_img = grayValue.astype(np.uint8)
    return gray_img

orig = cv2.imread(r'C:\Users\Jackson\Desktop\drum.png', 1)
g = grayConversion(orig)

cv2.imshow("Original", orig)
cv2.imshow("GrayScale", g)
cv2.waitKey(0)
cv2.destroyAllWindows()

answered Jul 11, 2018 at 13:49

Jeru Luke

21.5k13 gold badges85 silver badges91 bronze badges

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

Comments

Eric · Accepted Answer · 2018-07-11 13:39:54Z

2

You can use a dot product:

gray_image = image.dot([0.07, 0.72, 0.21])

Or even just do the whole operation manually:

b = image[..., 0]
g = image[..., 1]
r = image[..., 2]
gray_image = 0.21 * r + 0.72 * g + 0.07 * b

Don't forget to convert back to 0-255:

gray_image = np.min(gray_image, 255).astype(np.uint8)

edited Jul 11, 2018 at 13:39

answered Jul 11, 2018 at 13:36

Eric

98.1k54 gold badges257 silver badges389 bronze badges

2 Comments

thesamiroli Over a year ago

The image is still a 3-dimensional array.

Eric Over a year ago

You mean gray_image.ndim == 3? It's 2 on my machine

v.tralala · Accepted Answer · 2022-12-16 09:07:43Z

2

Solution using `apply_along_axis`

A solution can be achieved by using apply_along_axis:

import numpy as np
def grayscale(colors):
    """Return grayscale of given color."""
    r, g, b = colors
    return 0.21 * r + 0.72 * g + 0.07 * b

image = np.random.uniform(255, size=(10,10,3))
result = np.apply_along_axis(grayscale, 2, image)

Examples

10x10 image

We can now proceed to visualise the results:

from matplotlib import pyplot as plt
plt.subplot(1,2,1)
plt.imshow(image)
plt.subplot(1,2,2)
plt.imshow(result, cmap='gray')

Textual example (2x2 image)

To visualise the actual results in text I will use a smaller array, just a 2x2 image:

image = np.random.uniform(250, size=(2,2,3))

The content is:

array([[[205.02229826, 109.56089703, 163.74868594],
    [ 11.13557763, 160.98463727, 195.0294515 ]],

   [[218.15273335,  84.94373737, 197.70228018],
    [ 75.8992683 , 224.49258788, 146.74468294]]])

Let's convert it to grayscale, using our custom function:

result = np.apply_along_axis(grayscale, 2, image)

And the output of the conversion is:

array([[127.62263079, 157.64461409],
   [117.94766108, 197.76399547]])

We can visualise this simple example too, using the same code as above:

Further suggestions

If you want to apply your own custom function, then apply_along_axis is the way to go, but you should consider using purer numpy approaches such as the one suggested by Eric or, if possible, just load the black and white image using cv2 option:

cv2.imread('smalltext.jpg',0)

edited Dec 16, 2022 at 9:07

v.tralala

1,7653 gold badges21 silver badges46 bronze badges

answered Jul 11, 2018 at 12:54

Luca Cappelletti

2,55522 silver badges36 bronze badges

6 Comments

thesamiroli Over a year ago

so, the result variable should be a grayscale image right ? or am i missing something because when I try to display it, it only shows white pixels everywhere.

Luca Cappelletti Over a year ago

Yes, but he asked for using his own custom function.

Eric Over a year ago

Your comment gives the misleading impression that this is better because it avoids a python loop - but a python loop is exactly what apply_along_axis does.

Luca Cappelletti Over a year ago

Eric Comment removed, you are right.

Luca Cappelletti Over a year ago

@thesamiroli I believe so. What does it happen when using the cv2 grayscale loading cv2.imread('smalltext.jpg',0)?

|

Collectives™ on Stack Overflow

Converting an image to grayscale using numpy

3 Answers 3

Comments

2 Comments

Solution using `apply_along_axis`

Examples

10x10 image

Textual example (2x2 image)

Further suggestions

6 Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

3 Answers 3

Comments

2 Comments

Solution using apply_along_axis

Examples

10x10 image

Textual example (2x2 image)

Further suggestions

6 Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related

Solution using `apply_along_axis`