Robust Zero-Watermarking for Color Images Using VGG16 and Discrete Wavelet Transform

Abstract

Classical watermarking systems have been created to protect images using spatial and transform domains. However, traditional watermarking systems are more vulnerable to various assaults. Deep learning-based watermarking has recently gained popularity for its contribution to image security. This work presents a zero-watermarking method for protecting color images using VGG16, the discrete wavelet transform (DWT), and chaotic encryption. DWT is applied to the input color image, and the resulting features are fed into the VGG16 pre-trained network for deep feature extraction. DWT decreases complexity by acting as a pre-feature extractor. The Henon chaotic map is applied to encrypt the binary watermark image. A zero watermark is produced by combining the derived image features with the encrypted binary watermark image via an exclusive OR operation. The experimental results show that our technique is effective and resilient against geometric and typical image-processing attacks.