An adaptive QIM- and SVD-based digital image watermarking scheme in the wavelet domain

This paper presents a blind, adaptive quantization index modulation (QIM)- and singular value decomposition (SVD)-based watermarking scheme to embed watermark bits in the approximation subband of the wavelet domain. The QIM technique adaptively determines the quantization step for each embedding block using a statistical model. The SVD technique uses the quantization step to modify the SVs of each embedding block. This modification ensures the SVs of highly textured blocks are largely modified and the SVs of smooth textured blocks are slightly modified. The successive packing interleaving (SPI) scheme and the one-way hashing functions are respectively applied to improve the robustness of the proposed system. A statistical-clustering method is also applied offline to decide two optimal weighting parameters for the QIM technique so the adaptive quantization steps are optimal for all embedding blocks. Experimental results demonstrate our scheme is robust against JPEG compressions down to a level of 20% quality factor. It also performs better than a peer SVD-based wavelet domain watermarking approach.