A Robust Digital Image Watermarking Methodology using an Ensemble-Based Classifier for Mitigation of Geometrical Attacks

Abstract

Digital watermarking is an effective approach to the problem of copyright protection, thus maintaining the security of digital products on the network. Recently, several machine learning based digital watermarking algorithms have been proposed. The major factors of the watermark image are sustainability and imperceptibility against geometrical attacks. The digital content was deformed by the processing of geometrical attacks such as cropping, sharing, and translation. The process of authentication of digital content is compromised in terms of quality and security. This paper proposed a machine learning-based ensemble classifier for the optimization of embedding coefficient. The optimized embedding coefficient increases the strength of robustness factors. The process of optimization focuses on two elementary parameters: correlation coefficient and PSNR value. The minimum value of the correlation coefficient of pixel value increases the compactness of the watermark image and the source image. The primary phase of the optimization-based watermarking algorithm is feature extraction. This paper applies discrete wavelet transform methods for feature extraction. The discrete wavelet transform is a rich texture feature component of raw images. The lower content of noise is optimized by machine learning algorithms. The proposed algorithm is simulated in MATLAB tools. For the validation of algorithms, we applied various standard watermarking images and estimated CC and PSNR. The experimental results outperform the existing methods of watermarking.