SSIM index: Clearly explained

Question

I am trying to understand the SSIM index to compare two images for similarity, but I am having difficulties understanding how it is done.

Therefore, I would like to ask, if someone can explain it step by step, possibly also using some images to visualize the process ?

That would be of great help !

What I have undertstood so far:

1. I take two pictures
2. I split picture 1 in N squared subpictures
3. I split picture 2 in N squared subpictures at the same location as in picture 1
4. I compute the SSIM index for each local subpicture