Blind watermarking algorithm combining NSCT, DWT, SVD, and HVS

Abstract In order to solve the contradiction between watermark transparency and robustness, this paper combines Non‐subsampled Contourlet Transformation (NSCT), Discrete Wavelet Transformation (DWT), Singular Value Decomposition (SVD), and Human Visual System (HVS) to improve a watermarking algorithm. Firstly, this algorithm performs a first‐level NSCT decomposition on the carrier image to obtain its low‐frequency part, and then performs a second‐level DWT on the low‐frequency part, extracts its second‐level low‐frequency sub bands and block these sub bands to sub‐blocks according to their size. Secondly, SVD is performed for each sub‐block and the best embedding position of the watermark according to the sensitivity of HVS for the average brightness and the modified cosine similarity measurement method is calculated. Finally, the watermark scrambled and encrypted by Arnold is embedded into the selected singular value. This algorithm can realize the blind extraction of watermark. Experimental results show that the algorithm can not only satisfy the transparency but the robustness of the watermark at the same time as well.