If an incoming print job is a PDF file and the printer is not a PDF printer, the file always goes through a PDF interpreter (Ghostscript, Poppler, or MuPDF) to get rasterized. This is needed when the file contains vector/high-level graphics, fonts, …, but is unneeded and can even produce bad interferences if each page is a simply a full-page bitmap (raster graphics).
There is even an official file format, PCLm, which is a subset of PDF but describes each page as full-page raster graphics. It is used for driverless printing on cheaper printers which do not provide a PDF interpreter. As the files are a subset of PDF, they are recognized as PDF format. Also certain PDF from scanning software (or hardware) are raster-only.
To speed up the filtering process and to avoid quality degradation caused by mis-matched resolutions we could recognize these special-case PDFs, extract the bitmaps from each page and convert them into PWG/CUPS Raster and then continue as we had gotten raster graphics in the first place.
We could let the pdftoraster/gstoraster filter (or already CUPSvia MIME types) identify whether the input is PCLm (or any other known raster-only PDF format) and in this case delegate to a pclmtoraster filter to extract the raster data from the pages and convert them into other raster formats, without PDF interpreter.
The student's task here is to find out about suitable raster-only PDF types (at least PCLm should be supported, the more file types, the better), make these files get identified, and write a CUPS filter which extracts the raster data and passes it on as CUPS Raster. The resulting code will get added to the cups-filters project.
Till Kamppeter, Project Leader OpenPrinting (till at linux dot com), TBD
C programming, CUPS
(L)GPL 2+, Apache 2.0, MIT