I needed to compare whether two PDF documents generated from the same TeX sources where different or not, since the generation method was slightly different for both.
Trying to compare it visually would be crazy, since the document contains more than 1400 pages. So I needed an automatic way to do it.
The way to automatically compare PDF pages was extremely easy: convert all pages to images, compare each pages and be warned only on the different ones.
After that, it only rests to check visually which differences contain those pages which have images that don’t match.