Design and Analysis of Digital Image Watermarking schemes using Machine Learning Algorithms

Abstract

The popularity of internet access makes the communication and distribution of multimedia newlinedata (images, video and audio) very easy. Due to this ease, multimedia data can be newlineduplicated and illegally distributed and thus violating the intellectual property rights. The newlineprotection of multimedia data has become a challenging issue in current scenario. In the last newlinefew years, digital watermarking has become the effective solution for copy right protection, newlinecopy protection and authentication. A number of image watermarking algorithms have newlinealready been developed by various researchers with distinctive applications. newlineThe main objective of this thesis is to design an imperceptible, robust and secure digital newlineimage watermarking schemes for the application of copyright protection, copy protection newlineand ownership assertion. Based on the domain, digital image watermarking schemes are newlinedivided into spatial and frequency domain schemes. In this thesis, the main focus is to build newlinethe frequency domain image watermarking schemes with machine learning algorithms to newlineincrease the imperceptibility and to enhance the robustness against image processing attacks. newlineIn the watermarking applications such as copyright protection, copy protection and newlineownership assertion, high degree of robustness is the basic requirement. The proposed newlineschemes in this thesis are based on fractional discrete Cosine transform (FrDCT) / subband newlineDCT (SB-DCT) / discrete wavelet transform (DWT) / lifting wavelet transform (LWT) newlinefollowed by singular value decomposition (SVD) / QR factorization that can be applied newlinesuccessfully to extract the features of selected image blocks. The blocks are selected using newlinethe statistical property of each image block or the texture analysis (based on fuzzy entropy) newlineof the blocks. The characteristics of human visual system (HVS) model are used to prioritize newlinethe blocks to be altered in order to achieve good visual quality...