Machine Learning Based Digital Image Watermarking

Abstract

Watermarking is the method of adding information in digital media for content protection and authentication. Digital image watermarking is one of the solutions to offer value added security on the top of authentication and data encryption for content protection in digital images. The watermarking methods can be frequential and spatial domain methods. In this work, the frequency domain based watermarking method is implemented with the use of discrete wavelet transform (DWT). Along with the frequency domain technique, here we are utilizing the combination of topical developments of the mathematical techniques and most advanced algorithms of machine learning for digital image watermarking process. The mathematical technique considered here is the principal component analysis (PCA) because of its property of dimensionality reduction which further enhances the robustness to watermarking process. The machine learning algorithm taken into account in this work is the support vector machine (SVM) algorithm to increase the accuracy of watermarking process.