Cannot influence encoding of generated pdf metadata (such as with -encoding arg)

Binary package hint: imagemagick

Kubuntu 10.04, imagemagick package version 7:6.5.7.8-1ubuntu1.

I have a document scanner (Canon P-150) which I use to scan all incoming bills, letters etc. I use scanimage (from sane-utils) to scan them into .tiff files. I then use convert to compress the images into 85% quality jpegs, and then finally into pdf files. I am Swedish, so I when I name the files they often contain international characters, notably the Scandinavian å ä and ö.

    $ convert -compress jpeg -quality 85 page1.tiff page2.tiff documentåäöÅÄÖ.pdf

The resulting pdf file has some automatically generated pdf metadata tags that I can't set; title and producer. They seem to be written using some naïve encoding that doesn't work well with my Swedish alphabet. An example of pdfinfo output from the "documentåäöÅÄÖ.pdf" document generated with the above command;

    Title: documentåäöÖ×Ì.pdf

This is also the title of the document when opened in Okular and other programs.

convert doesn't seem to listen to the -encoding argument in this case, so I cannot simply use "-encoding UTF-8" when generating the documents. (Nor can I set the metadata tags manually, which would be even nicer.)