I have some video recording of my screen where I am drawing some diagrams, but there are a lot of moments where the screen is idle. Because I am using Camtasia, the cursor is not saved directly on the video itself so it makes it easier to detect idle moments.

I extracted all frames using ffmpeg. Here are 3 sequential frames:

I tried many methods from numpy to opencv to phash but I am unable to get correct results for these images. They either return True for all or False for all, but as you can see only the first 2 images at the same, the last one is different.

Here is my latest code:

import cv2
import numpy as np
from skimage.metrics import structural_similarity as ssim

def mse(imageA, imageB):
    # Mean Squared Error between the two images
    err = np.sum((imageA.astype("float") - imageB.astype("float")) ** 2)
    err /= float(imageA.shape[0] * imageA.shape[1])
    return err

def are_images_equal(image1_path, image2_path, mse_threshold=100, ssim_threshold=0.95):
    try:
        # Read images
        image1 = cv2.imread(image1_path)
        image2 = cv2.imread(image2_path)

        if image1 is None or image2 is None:
            return False

        # Convert images to grayscale
        gray1 = cv2.cvtColor(image1, cv2.COLOR_BGR2GRAY)
        gray2 = cv2.cvtColor(image2, cv2.COLOR_BGR2GRAY)

        # Compute MSE
        mse_value = mse(gray1, gray2)
        if mse_value > mse_threshold:
            return False

        # Compute SSIM
        ssim_index, _ = ssim(gray1, gray2, full=True)
        if ssim_index < ssim_threshold:
            return False

        return True
    except Exception as e:
        print(f"An error occurred: {e}")
        return False