Robust Image Watermarking Based on Hybrid Transform and Position-Adaptive Selection

Haiqun Liu, Yi Chen, Guowei Shen, Chun Guo, Yunhe Cui

Abstract: Lots of information-hiding-based image watermarking methods have been widely used in the field of copyright protection. In these methods, researchers usually introduce optimization algorithms to improve the robustness and imperceptibility of watermarked images. However, conventional optimization algorithms often treat the entire image as a single optimization target and neglect the feature differences between different blocks in the image. Therefore, these unused differences have limited the improvement of watermarking performance. To address this issue, we propose a robust image watermarking method based on hybrid transforms and position adaptive selection in this paper. In this robust image watermarking method, an image is firstly decomposed into a directional sub-band by Non-Subsampled Contourlet Transform and Contourlet Transform. The directional sub-band is divided into many blocks and then each block performs a Discrete Cosine Transform (DCT) to obtain the DCT coefficient matrix. Next, we propose an optimization strategy by considering the feature differences to combine with the Cuckoo Search Algorithm to independently and adaptively select the optimal watermark embedding position for each DCT coefficient matrix. Finally, the watermark bit is embedded into the DCT coefficient corresponding to the selected position of each coefficient matrix. We conducted extensive experiments to verify the watermarking performance of our proposed method, experimental results have shown that the PSNR value of the watermarked image increases by 17% on average, the NC value between the watermark extracted after the attack and the original watermark is increased by 1% on average, and our proposed robust watermarking method has a good improvement in both the robustness and the imperceptibility when compared with related methods.