Secure and flexible image watermarking using IWT, SVD, and chaos models for robustness and imperceptibility

The utilization of Singular Value Decomposition (SVD) is widespread in several image watermarking methods to attain a more favorable balance between the demands of watermarking. Nevertheless, SVD exhibits some limitations, including false positives, limited capacity, and security issues. In order to overcome these challenges, a new approach to image watermarking has been introduced, which utilizes a combination of Integer Wavelet Transform (IWT), SVD, and chaos models. The generation of chaotic orbits in the scheme is based on utilizing chaotic variables generated through a secret key. The secret key is a major element in embedding, extraction, encryption, and ownership protection. The smooth generation of multiple scaling factors (MSF) matrices is achieved using chaotic points in the encryption stage. In order to attain a suitable equilibrium between watermarking robustness, imperceptibility, and security, the IWT technique is utilized to convert both the host and watermark images into frequency coefficients. The proposed method involves the extraction of a 4-bit matrix from the $$ sub-bands of the host image. This matrix is subsequently subjected to scrambling and merged with the watermark image, resulting in a 12-bit watermark image. Afterwards, the two sub-bands of the host image undergo two levels of IWT, and three sub-bands are chosen from the resultant 2-level image for embedding. The selected sub-bands are then subjected to SVD, where the embedding process utilizes the singular values $$ . The high-frequency sub-bands of the watermark image are embedded into the singular values $$ of the host image using chaotic MSF matrices. The scheme also encrypts the side information to achieve high security and protect the ownership. Even a slight change in the secret key can highly affect the proposed scheme, resulting in the generation of a noise image if an incorrect key is used. This feature makes the proposed scheme robust against false positive problem (FPP) attacks. Additionally, the proposed scheme can embed watermark images with high flexibility and low distortion, achieving high robustness and imperceptibility under different hybrid MSF matrices, outperforming several other schemes. With these features, the proposed scheme successfully fulfils the security and capacity requirements of a blind watermarking scheme, making it suitable for practical watermarking applications.