Dual-color blind image watermarking algorithm using the graph-based transform in the stationary wavelet transform domain

Abstract

Aiming at the problem of color image copyright protection, a dual-color blind image watermarking algorithm using the graph-based transform (GBT) in the stationary wavelet transform domain (SWT) is proposed. First, the color watermark image and the color carrier image are separated into their R, G, and B channel components. Second, the R, G, and B components of the color watermark are scrambled by the Fibonacci–Lucas transform and encrypted with the cosine-transform-based chaotic system (CTBCS). Each component of the encrypted watermark is then converted into a binary sequence. Third, the low-frequency sub-bands from each component of the carrier image performed by the SWT are divided into non-overlapping blocks. Finally, the three binary watermark sequences are embedded into the GBT low-frequency coefficients of selected image blocks of the corresponding component by the parity quantization. Before the watermark extraction, the image to be authenticated is rotationally corrected based on the Fan-beam transform. This operation improves the resistance of the proposed algorithm to rotation geometry attacks. Experimental results show that the algorithm is robust to conventional signal processing attacks and to arbitrary angle rotation attacks while effectively ensuring the invisibility of the watermark.