Python & OpenCV - Trying to turn transparent pixels into white nontransparent pixels

Question

I'm relatively new to data processing using OpenCV. For a machine learning project, I'm trying to process images so that all the transparent pixels will be turned into white nontransparent pixels for my GAN to use. However, I keep on getting an error when trying to process it. Here is my code below /

    img = cv2.imread(dirfName, cv2.IMREAD_UNCHANGED) (Note Dirftname is a defined variable)
    for col in img:                        
        for row in img[col]:                
            if(img[col,row] == (0,0,0,0)):                    
                    img[col,row] = (255,255,255,255)

    cv2.imwrite((fCropPath + fNameIndent + fname), new_img)

The error is with the if statement and it states: ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()

Answer 1

Note your syntax in the itterations, row and col are not indexes.

import cv2
import numpy as np

img = cv2.imread(dirfName, cv2.IMREAD_UNCHANGED)
for col in img:
    for row in col:
        if(row.__eq__(0).all()): # This is what you want
                row += 255

cv2.imwrite((fCropPath + fNameIndent + fname), new_img)

row.__eq__(0).all() is equivalent to np.all(row == 0)

Furthermore, it is better practice to vectorize your loops:

import cv2
import numpy as np

img = cv2.imread(dirfName, cv2.IMREAD_UNCHANGED)
ind = np.equal(img, 0).all(axis=2)
img[ind] = np.array([255]*4)

cv2.imwrite((fCropPath + fNameIndent + fname), new_img)

The above code does exactly the same thing but vectorized.