Design of Robust Content Based Image and Video Watermarking Schemes and Architecture in Compressed Domain

Abstract

Over the last decade, with the rapid growth of the internet and digital media techniques, multimedia data such as images and videos can easily be copied, altered, and distributed over the network without any loss in quality. Therefore, protecting the ownership of multimedia data has become a very significant and challenging issue. newlineThis thesis is dedicated to the design of robust content-based image and video watermarking techniques for copyright protection. This study introduces blind, robust image watermarking techniques that achieves imperceptibility, resilience, security, tampered area detection, and recovery. Video watermarking methods are proposed to boost resilience and visual quality of watermarked video. To improve efficiency key-frame based video watermarking approaches are developed. Simulink based prototyping architectures for image and video watermarking techniques are presented to provide graphical environment for modeling and simulating the watermarking systems. newlineAs regards to content based image watermarking for copyright protection, we have developed two image watermarking approaches: newlineIn the first approach, we have proposed a robust and blind image watermarking scheme using DCT and SURF, wherein the SURF feature point based Square regions are detected. Watermark is embedded in the selected region using DCT to enhance perceptual quality of watermarked image and resistance against several attacks. The presented technique outperforms the existing methods on standard image dataset (University of Southern California-Signal and Image Processing Institute) in terms of visual quality and robustness. Experimental results reveal that the stability of SURF feature points improves the robustness of proposed watermarking technique. newlineIn the second approach, a new self-embedding image authentication technique for tamper detection and recovery using SVD and DCT is presented. Initially, the original image is divided into non-overlapping blocks which are then separated into two sets using SVD. Thereafter, DCT and SVD