A Tree-Structure Based Logo Watermarking Scheme

A novel wavelet tree-structure based logo watermarking scheme for gray-scale image is proposed. Based on the same hierarchical idea in embedded zerotree wavelet (EZW) compression, a discrete wavelet transform (DWT) decomposed host image can be viewed as a collection of coefficient trees whose parents are coefficients in the approximate subband (LL). The human vision system (HVS) characteristics of the meaningful logo image are exploited by the same level DWT to generate a similar tree structure. Each logo tree and its LL parent are embedded separately and repeatedly into the perceptually significant trees and LL coefficients of the host image with a well-known perceptual model. By grouping coefficient trees of the host image with different sizes, we introduce two embedding approaches under the same scheme, the pixelwise approach and the blockwise approach. Aiming to feasible hardware implementation, experimental results using the integer-version of DWT are presented to verify the performance of the scheme and its robustness against many common attacks