Robust Color Image Watermarking Using Denoising Autoencoders: A Novel Approach for Digital Rights Management

Abstract

In this paper, a unique method for colour image watermarking through Denoising Autoencoders (DAEs) is proposed. In this paper, we propose a deep learning-based watermarking solution to enable secure embedding and extraction of watermarks in colour images that achieves robustness against image quality preservation attack. Our approach employs a specially designed DAE structure to acquire lossy embedding and extraction that is both imperceptible and robust, respectively. We conduct extensive experiments with a large dataset containing 5000 diverse colour images to show the superiority of our method over traditional watermarking techniques and other deep learning-based approaches. Experimental results prove that the proposed scheme has a superior robustness in parallel with common image processing and geometric attacks, rendering it suitable for real-world DRM applications.