Python - Numpy, Pillow, OpenCV - compare images 1

Make a simple .py which compares two images and returns the percentage of similarity. Imagine I draw a line in "Image B" which is a copy of "Image A". Now when we compare the both, it should return say 97% similar. If I draw more lines in "Image B" it should return 89% similar etc. The process should take no more that 5 seconds. You can make the image grayscale or reduce the size even more. Here you can find two sample images in large scale and small scale: [login to view URL]