opencv python merge different channel images into one

Question

I have few satellite images each of them represents one channel of main satellite image, these are 11 images in total, each are labled with different channel, all images are in .tiff format with grayscale colorspace, now i simply want's to merge these images into one, to represent all channels into one image, so is this possible, remeber here, I don't want to concat images , which can be done using this:

vis = np.concatenate((img1, img2), axis=1)

I want to merge all of them into one single image , without distorting the data contained within, few channel images are attached below.

Any help is appreciated.

Solution 1

As OpenCV 3.x stored image as numpy array, we can simply average each image and add them together, provided that the height and width of the images are exactly the same.

img_1 = cv2.imread('./imagesStackoverflow/sat_1_331-442.png')
img_2 = cv2.imread('./imagesStackoverflow/sat_2_331-442.png')
img_3 = cv2.imread('./imagesStackoverflow/sat_3_331-442.png')
img_4 = cv2.imread('./imagesStackoverflow/sat_4_331-442.png')

no_img = 4
img = img_1/no_img + img_2/no_img + img_3/no_img + img_4/no_img

To get a quick result, I manually edited the size of the four images to 442(h) x 331(w) pixels.

Below is the merged image, with 3 channels.

To merge 11 images, you may just extend the code as:

img = img_1/no_img + img_2/no_img + img_3/no_img + ... + img_11/no_img

Solution 2

First, you need to carefully consider the number of channels that you have so that you can create a useful image. In the example below I assume that you have three channels (Red, Green, and Blue) which could be combined into an RGB image.

import numpy as np
import cv2

"""Read each channel into a numpy array. 
Of course your data set may not be images yet.
So just load them into three different numpy arrays as neccessary"""
a = cv2.imread('chanel_1.jpg', 0)
b = cv2.imread('chanel_2.jpg', 0)
c = cv2.imread('chanel_3.jpg', 0)

"""Create a blank image that has three channels 
and the same number of pixels as your original input"""
needed_multi_channel_img = np.zeros((a.shape[0], a.shape[1], 3))

"""Add the channels to the needed image one by one"""
needed_multi_channel_img [:,:,0] = a
needed_multi_channel_img [:,:,1] = b
needed_multi_channel_img [:,:,2] = c

"""Save the needed multi channel image"""
cv2.imwrite('needed_multi_channel_img.png',needed_multi_channel_img)

Solution 3

I've been trying this for 2 days straight and one thing that I failed to understand was that the merge function of opencv takes a tuple, not 3 different arguments.

image = cv2.imread(image_file, 1)
R, G, B = cv2.split(image)
output_image = cv2.merge ( (R, G, B) )

Solution 4

What do you mean by merge ? They are all grayscale images, so they are basically the same channel in grayscale color space. Some information are doomed to be lost when you try to blend or add the images. Try the above two methods using functions in OpenCV or Pillow.

mul1 = ImageChops.add(img1, img2, scale=2)
mul2 = ImageChops.add(img3, img4, scale = 2)
mul3 = ImageChops.add(mul1, mul2, scale = 2)

mul3.show()

R1 = img1.convert('RGBA').resize([456,512])
R2 = img2.convert('RGBA').resize([456,512])
R3 = img3.convert('RGBA')
R4 = img4.convert('RGBA')

S1 = ImageChops.blend(R1,R2,0.5)
S2 = ImageChops.blend(R3,R4,0.5)
S3 = ImageChops.blend(S1,S2,0.5)
S3.show()

View more solutions