Automated Extraction of Paper Watermarks

The aim of this paper is to outline the issues involved in the application of machine vision to the automatic extraction of watermarks from paper samples. The correct identi cation and localisation of watermarks is a key issue in paper manufacturing. As well as requiring the position of the watermark for defect detection and classi cation, it is necessary to insure its position on the paper prior to the cutting process. Two paper types are discussed, with and without laid and chain lines (these lines appear as a complex periodic background to the watermark and further complicate the segmentation process). We will examine both morphological and Fourier approaches to the watermark segmentation process, concentrating speci cally on those images with complex backgrounds. Finally we will compare and contrast these approaches.