CAM::PDF (and likely others) can extract images, which doesn't mean that that they necessarily have the desired metadata embedded in them, that depends on the tools used to create the jpg and the PDF, and the options selected.