Pseudo-noise Patterns for Watermarking

This paper presents an overview of pseudonoise patterns suitable for embedding in images for the purposes of information hiding, including some new constructions. Typically, correlative recovery is used in such schemes. In order to minimize errors, a single pattern embedded in an image should possess a solitary autocorrelation peak and low off-peak values. Cross-correlation with the image is reduced by using a pattern of suitable size. It is possible to embed multiple patterns, which have greater information storage capacity and can be made to appear more random, or resistant to cryptographic attack. Such patterns must also have low cross-correlation with each other and are thus more difficult to construct. Still images can contain twodimensional patterns, whilst video can accommodate threedimensional arrays. Currently, all constructions of such arrays are based on one-dimensional pseudonoise sequences, whose fundamentals have been derived from communications and radar.