How to compress an image with (basic) linear algebra