Robust zero-watermarking algorithm for digital color image based on NasNet-Large and modular integrated logistic exponential map
摘要
In the zero-watermarking approaches for images, a watermark sequence has a logical association with the source image rather than being physically embedded. Zero-watermarking approaches have become an important area of research in the digital watermarking field due to their efficacy and distortion-free nature. This paper presents a novel, robust, and secure algorithm for zero-watermarking color images using a neural architecture search network-large (NasNet-Large (and modular integrated logistic exponential map (MILEM). First, we used the pre-trained NasNet-Large to extract deep feature maps from the host color image. An XOR operation is carried out between the owner’s watermark sequence and the binary feature representation obtained from the deep feature maps. Finally, to guarantee security, MILEM scrambles the binary feature matrix and encrypts the copyright watermark. The testing findings showed that the suggested algorithm maintained exceptional robustness against diverse attacks, achieving a bit error rate (BER) below