An Optimized Color Image Watermarking Scheme based on HD and SVD in DWT Domain

Abstract

Digital watermarking is considered a trustworthy strategy for proving ownership of valuable digital files such as audio, image, and video documents. Most of the prevailing image watermarking systems embed grayscale or binary image watermarks, while only a few use color images as watermarks. In this paper, we develop a secure, imperceptible, and robust optimized semi-blind color image watermarking technique that uses color images as watermarks. It is based on Hessenberg Decomposition (HD) and Singular Value Decomposition (SVD) in the Discrete Wavelet Transform (DWT) domain. First, the color host image and color watermark image in RGB space are converted to YCbCr space, and then the watermark data are embedded into the luminance component (Y) of the host image. In this work, the principal component of the watermark's luminance (Y) is implanted into the associated singular value of the host image with an appropriate scaling factor that optimizes the robustness-imperceptibility tradeoff. The Artificial Bee Colony (ABC) algorithm is used to find the appropriate scaling factors. To further enhance the security, the Arnold transformation is used to scramble the Y channel of the watermark before it is injected into the host image. As demonstrated by the Peak Signal to Noise Ratio (PSNR) and Normalized Correlation (NC) metrics, the proposed scheme exhibits high invisibility and is robust to most image processing manipulations, geometric operations, and combinational attacks. Compared to various existing color image watermarking schemes that use color images as watermarks, it shows higher imperceptibility and robustness.