Introduces a preprocessing that scans each page for page sized images.
If one is encountered, all images that are below a configured ratio in
respect to the page size are dropped.
This step has to occur before the image stiching logic, but MIGHT
introduce the problem of dropping image parts that might constitue an
image. This hoever is not solveable since we want to drop the small
images before further processing since the faulty character images are
also stiched to a valid image, that in reality isn't an image.
4102a564a3