Reversible watermarking in compressed images

Abstract

Reversible Watermarking schemes are used to embed ownership information into video or images and are able to restore the original image after watermark extraction. However, these will fail under lossy compression due to data truncation. The purpose of this project is to create a reversible watermarking scheme suitable for JPEG compression. Unlike conventional reversible watermarking schemes used in compressed domain, which are generally based on manipulating the AC components of the quantized DCT domain produced by the JPEG codec. The proposed reversible watermarking algorithm exploits and manipulates both the AC and DC components. The proposed scheme uses the prediction expansion approach for watermark embedding and is able to extract the watermark without requiring the original image thus making it a blind reversible watermarking scheme. The simulation results performed demonstrate the superiority of the proposed scheme to recently published methods, where the proposed scheme is able to embed a large watermark and introduce only a fair amount of distortion into the image quality compared to these methods.