merge two images with alpha channel

Question

I have two images, one with and other without alpha channel. Thus, image A and B has a shape of (x,y,4) and (x,y,3) respectively.

I want to merge both images in a single tensor using python, where B is the background and A is the upper image. The final image must have a shape of (x, y, 3). I tried if scikit-image or cv2 is capable of doing this, but I couldn't found any solution.

here is a link for alpha blending using matlab https://github.com/techmn/ImageBlending — user8190410, Apr 09 '18 at 16:46
I want to overlap one over the other, where the upper image is partially transparent. — nunodsousa, Apr 09 '18 at 16:52
Does this answer your question? [How to merge one RGBA and one RGB images in opencv](https://stackoverflow.com/questions/70578600/how-to-merge-one-rgba-and-one-rgb-images-in-opencv) — Thomas Weller, Jan 04 '22 at 13:57

user8190410 · Accepted Answer · 2018-11-10T14:55:09.720

7

here is alpha blending in python

import numpy as np
import cv2

alpha = 0.4

img1 = cv2.imread('Desert.jpg')
img2 = cv2.imread('Penguins.jpg')

#r,c,z = img1.shape

out_img = np.zeros(img1.shape,dtype=img1.dtype)
out_img[:,:,:] = (alpha * img1[:,:,:]) + ((1-alpha) * img2[:,:,:])
'''
# if want to loop over the whole image
for y in range(r):
    for x in range(c):
        out_img[y,x,0] = (alpha * img1[y,x,0]) + ((1-alpha) * img2[y,x,0])
        out_img[y,x,1] = (alpha * img1[y,x,1]) + ((1-alpha) * img2[y,x,1])
        out_img[y,x,2] = (alpha * img1[y,x,2]) + ((1-alpha) * img2[y,x,2])
'''

cv2.imshow('Output',out_img)
cv2.waitKey(0)

edited Nov 10 '18 at 14:55

answered Apr 09 '18 at 17:03

user8190410

1,264
2
8
15

1

How does a JPG have ARGB color? This does not answer the question, IMHO. Don't know how this can be the accepted answer. It does not use the alpha channel of the upper picture. – Thomas Weller Jan 04 '22 at 12:33
Kindly check what is alpha blending first https://docs.opencv.org/3.4/d5/dc4/tutorial_adding_images.html . The answer does not say that it uses alpha channel. – user8190410 Jan 04 '22 at 12:50
1

OP: "I have two images, one with and other without alpha channel" What is (x,y,4) if not RGBA? – Thomas Weller Jan 04 '22 at 12:55
If you are using png images and want to use alpha mask then you can see this tutorial https://learnopencv.com/alpha-blending-using-opencv-cpp-python/ – user8190410 Jan 04 '22 at 12:56
Also, if you don't want transparency in your final image you can simply delete the 4th / alpha channel – user8190410 Jan 04 '22 at 13:03

score 0 · Answer 2 · answered Apr 10 '18 at 07:52

0

The above solution works, however I have a more efficient one:

alpha = A[:,:,3]
A1 = A[:,:,:3]

C = np.multiply(A1, alpha.reshape(x,y,1)) + np.multiply(B, 1-alpha.reshape(x,y,1))

answered Apr 10 '18 at 07:52

nunodsousa

2,635
4
27
49

merge two images with alpha channel

2 Answers2