A canvas is a coordinate space on which images can be painted. Images are painted onto a canvas through annotation. In the simple case we're only painting a single image annotation so that it fills the whole canvas.
We call it a shared canvas because it allows for painting multiple images onto the same canvas as well as overlaying additional annotations.
Think of a canvas as an empty presentation slide. You can put a page image on the canvas via annotation:
Once you have a base image painted onto the canvas you can then apply further annotations like transcriptions, translations, scholarly commentary.
While just one related piece of how IIIF works, the shared canvas model is vital for understanding what IIIF enables and how.