A Robust Watermarking Scheme Based on Dual Quantization of Wavelet Significant Difference

Abstract

This paper proposes a blind robust watermarking algorithm based on the quantization of significant DWT coefficients. Low frequent wavelet coefficients of the host image are randomly permutated into sub-groups firstly. To guarantee watermark robustness while preserving good fidelity, watermarking modifications are distributed on the maximum positive and minimum negative coefficients of each group which preserve the most significant local amplitudes. The difference between two largest positive (or smallest negative) coefficients is quantized to an even or odd multiple of a quantization step parameter Q according to the watermark bit to be embedded. The blind watermark extraction could be straightforwardly achieved by checking the parity of the quotient between significant differences and Q. Comparison experiments with existing methods demonstrate the superiority of our scheme on robustness against content-preserving operations and incidental distortions such as JPEG compression, Gaussian noise.