Ok. I tried a program called PDF-Analyzer now, which says there’s 40 embedded objects, 9 embedded image files, and two document layers. There are also two compression modes used: Flate and DCT (jpeg?). If we look at the images that inkscape creates on the harddisk if you follow the steps paulycy outlined, we note:
- A greenish image that contains some signatures, a little white background where the text should be, some black spots here and there.
- A mostly black image that contains mostly white text.
- The state registrar stamp at the bottom, and his signature, again white on black.
- The date stamp, to the left of it, white on black.
- Some other small parts of text, or just spots.
Now these, together with the different compression modes, seem to indicate that the PDF uses so-called MRC-compression: The scanned image is automatically seperated into two layers, one of which is text, one is background, which are compressed using different algorithms. This is a very effective method for PDF compression that makes intelligent use of the capabilities of the PDF file format (it would be even better if they used JBIG2 and JPEG2000 instead of Flate and DCT, which PDF also supports). Except that this particular MRC-scheme apparently uses only one layer for the background, and adds the text parts as objects.
I agree with D-fendr: This is most likely software generated.