The computer-aided detection of inferior printing quality and errors

This paper describes a new approach which detects inferior printing quality and errors by using a regular PC and document scanner. Our method relies on the comparison of an inspected document with its referential version. It firstly registers the images of the two documents, and then detects any discrepancies between the aligned pixels or regions. Iterations of the two-step registrations with interim interpolations introduce a sort of elastic image correction. We confirmed experimentally that the error detection rate for those documents with simpler structures, mostly pictures, was about 95%, whereas with more complex documents containing a lot of text this figure is about 90%