Here are a couple of techniques which may not be in use anymore.

"Rook move encoding" for outlines of shapes, in particular digitized text characters. Instead of recording all the black and white pixels inside the character's body, record the outline of the character, in xy pixel space, as if you were moving a chess rook. E.g. 5 up, 2 left, 4 up, 3 left, ..... I worked with laser printers which did this.

"Patchified memory". Instead of treating a page as a set of raster lines, treat it as a set of small squares, like a quilt. Printed pages of text + black/white graphics tend to have lots of such patches which are all-white or all-black. Those patches can be represented in a very compact way.

These techniques are mathematically unsophisticated, but are easy to understand. Oh, also, they are both "lossless".