Adding a same value to each channel in a 3 channel image?

Question

I read an image :

img = cv2.imread("sky.png", 1)

Now, I want to add one pixel in each column for each channel. The way I attempted this is as follows:

img[row_1, :, 0] = np.insert(img[row_1,:,0], column_1, some_value)
img[row_1, :, 1] = np.insert(img[row_1,:,1], column_1, some_value)
img[row_1, :, 2] = np.insert(img[row_1,:,2], column_1, some_value)

Is there a better way of doing this, than to write to each channel separately?

UPDATE: As I have mentioned, that I want to add a new column, that is a 4x4 image, gets transformed to 4x5 image. The value of each pixel is different and the order of columns is also not fixed. For example, first pixel gets inserted into 3 column and second pixel gets inserted into 1 column and so on (using a predetermined set of columns)

Example:

[
  [1,2,3,4],
  [4,5,6,7],
  [8,9,10,11]
]

The above is a 3x4 image (will be a 3 channel image, in reality). I want to convert this to 3x5 image by adding pixels at [0,2] , [1,1], [2,4]

The output then becomes:

[
   [1,2, new-pixel-a, 3, 4],
   [4,new-pixel-b, 5, 6, 7]
   [8, 9, 10, 11, new-pixel-c]
]

So I get a new image, (3, 5)

Answer 1

Looking at the documentation on NumPy's insert method, I would come up with the following solution:

import cv2
import numpy as np

# Read image; output image dimensions
image = cv2.imread('N8e9S.png')
print(image.shape)

# Set up column indices where to add pixels
colIdx = np.array(image.shape[0] * np.random.rand(image.shape[0]), dtype=np.int32)

# Set up pixel values to add
pixels = np.uint8(255 * np.random.rand(image.shape[0], 3))

# Initialize separate image
newImage = np.zeros((image.shape[0], image.shape[1]+1, 3), np.uint8)

# Insert pixels at predefined locations
for i in range(colIdx.shape[0]):
    newImage[i, :, :] = np.insert(image[i, :, :], colIdx[i], pixels[i, :], axis=0)

# Output (new) image dimensions
print(newImage.shape)

# Show final image
cv2.imshow('image', image)
cv2.imshow('newImage', newImage)
cv2.waitKey(0)
cv2.destroyAllWindows()

Input image is this one:

The final output looks like this:

Print output to verify new image dimensions:

(241, 300, 3)
(241, 301, 3)

Hope that helps!