This document proposes a half-toning-based multilayer watermarking technique of low computational complexity. Multiple watermarks are embedded into a single watermark for improved security and embedding capacity. Reference tables are generated using direct binary search to ensure the output is in halftone format. Watermarks are then embedded through table lookup using optimized compressed tables with various textural angles. The decoder increases differences among embedding angles and reduces required dimensions using least mean square metric. A naïve Bayes classifier is used to extract the embedded watermarks by collecting possibilities of multilayer information to classify the associated angles. Experimental results show the method requires only 8.4 ms for embedding and 2 MB of storage space.

1. PROGRESSIVE HALFTONE WATERMARKING USING MULTILAYER TABLE
LOOKUP STRATEGY
ABSTRACT
In this paper, a half-toning-based multilayerwatermarking of low computational
complexity is proposed.An additional data-hiding technique is also employed to embedmultiple
watermarks into the watermark to be embedded toimprove the security and embedding capacity.
At the encoder,the efficient direct binary search method is employed togenerate 256 reference
tables to ensure the output is in halftoneformat. Subsequently, watermarks are embedded by a set
ofoptimized compressed tables with various textural angles fortable lookup. At the decoder, the
least mean square metric isconsidered to increase the differences among those
generatedphenotypes of the embedding angles and reduce the requirednumber of dimensions for
each angle. Finally, the naïve Bayesclassifier is employed to collect the possibilities of
multilayerinformation for classifying the associated angles to extract theembedded watermarks.
These decoded watermarks can befurther overlapped for retrieving the additional hidden-
layerwatermarks. Experimental results show that the proposedmethod requires only 8.4 ms for
embedding a watermarkinto an image of size 512 × 512, under the 32-bit Windows 7platform
running on 4GB RAM, Intel core i7 Sandy Bridgewith 4GB RAM and IDE Visual Studio 2010.
Finally, only 2 MBis required to store the proposed compressed reference table.