Difference of adding two images using NumPy and OpenCV

Recipe Objective: What is the difference between adding two images using NumPy and using OpenCV?

In this recipe, let us discuss the difference between adding two images using NumPy and using OpenCV.

Get Access to Plant Species Identification Project using Machine Learning

Step 1: Import libraries and read the images

The two images that we are using in this recipe are given below.

import cv2
import numpy as np
image1 = cv2.imread('project.jpg')
image2=cv2.imread('OpenCV_Logo.jpg')


Step 2: Compare and alter the sizes

Images with different sizes can not be added, and hence we compare the sizes and resize one of the images.

print(f"The shape of image 1 and image 2 is {image1.shape} and {image2.shape} respectively")

Output:

The shape of image 1 and image 2 is (110, 335, 3) and (190, 340, 3) respectively

Hence we reshape the first image as the shape of the second image using the np.resize() function.

image1=np.resize(image1,image2.shape)

Step 3: Numpy addition vs OpenCV addition

We already know that the pixel arrays are stored in uint8 format. uint8 is nothing but an unsigned integer of 8 bits length. These generally hold values in the range of 0 to 255. The main difference between NumPy and OpenCV addition is that NumPy addition is based on modulo operation while OpenCV uses saturated operation. Let us break down the above statement with an example.

When we add two uint8 numbers using NumPy, we will get a modulus of the sum if the sum exceeds 255

For instance : 170 + 150 = 320 % 256 = 64

Here 320 is the actual sum, and since it is more significant than 255, 320 mod 256 is calculated, which gives 64

But when we add two uint8 numbers using the cv2.add() function, we would get the result as 255 if the sum exceeds 255.

For instance : 170 + 150 = 320 => 255

Here 320 is the actual sum, and 255 is the final result after saturation operation.

And hence, for this reason, it is always advisable to use cv2.add() function too add images over normal NumPy addition

numpy_addition = image1 + image2
cv2_addition = cv2.add(image1,image2)