Preprocessing Images for OCR

Hi!

How do you preprocess images before using
the OCR?

Currently I use convert from imagemagick like
this:
 convert -normalize -blur 1 -threshold 50% orig.png ocr.bmp

The scanned pictures are cards which where filled
out by hand. The only contain numbers and the numbers are
written into a line with boxes in the background.

Is there someone who wants to share his knowledge?

I use kadmos on linux.

 thomas