Here's an idea for a workaround: if you were to do OCR on an image, the resulting text can be stored in the pdf, with the characters lining up. How is this done? You could probably then have dummy text behind each image, just to give the software something to select and attach a journal to.