Hands-On Genetic Algorithms with Python

Pixel-based Mean Squared Error

The most common way to evaluate the similarity between images is by conducting a pixel-by-pixel comparison. This requires, of course, that both images are of the same dimensions. The MSE metric can be calculated as follows:

1. Calculate the square of the difference between each pair of matching pixels from both images. Since each pixel in the drawing is represented using three separate values – red, green and blue – the difference for each pixel is calculated across these three dimensions.
2. Compute the sum of all these squares.
3. Divide the sum by the total number of pixels.

When both images are represented using the OpenCV (cv2) library, this calculation can be done in a very straightforward way, as follows: