python - How to add images on top of each other in for loop?

Question

I've got 4 images. I want to get 1 image which is an overlay of all 4 images. My code doesn't do this correctly. Here is my code so far:

framefiles = [file for file in os.listdir(inputvideopath) if os.path.isfile(os.path.join(inputvideopath, file)) and file.endswith('jpg')]
for frame_id, frame in enumerate(framefiles):
    if frame_id < 1:
        img1 = cv2.imread(output_dir + frame)
    if frame_id >= 1:
        img2 = cv2.imread(output_dir + frame)
        final_img = cv2.add(img1,img2)
        cv2.imshow('hh',final_img)
        cv2.waitKey(0)

Any help would be greatly appreciated. Thank you.

question from:https://stackoverflow.com/questions/65936683/how-to-add-images-on-top-of-each-other-in-for-loop

Answer 1

You better DON'T use loops, if you don't want to handle exact weights. With each iteration you'd give the newly overlayed image proportionately too much weight w.r.t. to the already merged image.

Let's see these four images:

I prepared some code to visualize:

import cv2
import numpy as np

images = ['1.png', '2.png', '3.jpg', '4.jpg']
images = [cv2.resize(cv2.imread(i), (400, 400)) for i in images]

# Don't do loops, m'kay?
output = images[0]
for i, image in enumerate(images[1:]):
    output = cv2.addWeighted(output, 0.5, image, 0.5, 0)
    cv2.imwrite(str(i) + '.png', output)

# Do linear blending using all images at once.
output = (np.array(images) / len(images)).sum(axis=0).astype(np.uint8)
cv2.imwrite('output.png', output)

The (intermediate) output(s) of the loop look like that:

The night image is most prominent, whereas Paddington can be hardly seen.

It'd be better to divide all images by the number of images you want to overlay, and then sum them. Using NumPy, that's the given one-liner, and that'd be the output:

----------------------------------------
System information
----------------------------------------
Platform:      Windows-10-10.0.16299-SP0
Python:        3.8.5
NumPy:         1.19.5
OpenCV:        4.5.1
----------------------------------------