Watermarking of Medical Images for Content Authentication and Copyright Protection

Abstract

Due to development of latest technologies in the areas of communication and computer networks, present businesses are moving to the digital world for effectiveness, convenience and security. One of the examples is in the area of medical applications, where traditional diagnosis is being replaced by e-diagnosis. This paradigm gave birth to number of applications in healthcare industry like teleconsulting, telesurgery and telediagnosis. All these applications require the exchange of medical images in digital format from one geographical location to another throughout the globe via a cheap and fast network such as Internet. However, digital form of medical image can easily be manipulated through image processing softwares. Insurance companies, hospitals as well as patients might want to modify the medical images for a number of reasons. The tampered images may be used for illegal purposes. The objective of this dissertation is to investigate how to protect a medical images, detect tampered images and recover the original image using the technique of Digital Watermarking. Watermarking tech- nology can be categorized into robust, fragile and semi-fragile watermarking, each serving for different purposes. In order to facilitate sharing and remote handling of medical images, we propose techniques to solve the problem of copyright protection and content authentication using robust and fragile watermarking respectively. In the first part of this dissertation, we design a fragile watermarking system for the exact authentication of medical images. The proposed system authenticates the computed tomography (CT) scan medical images of thorax area against different distortions. The system first enhances the embedding capacity of a CT scan image by isolating the actual lung parenchyma unlike classical watermarking techniques, which use logical square or ellipse for isolating region of interest (ROI). Secondly it embeds watermark only in region of non interest (RONI), thus not compromising the diagnosis value of medical image. The method utilizes the spatial domain watermarking and least significant bit (LSB) replacement method for embedding the watermark. Experimental results reveal that the proposed system detects both legitimate and illegitimate distortions. Robust watermarks are well suited for copyright protection because they stay intact with the image under various manipulations. In the second part of the dissertation, we investigate an approach of robust watermarking for protection of medical images which ensures the security of medical images against geomet- rical distortions. In the proposed method we firstly normalize the medical image using image moments in order to make it invariant against geometric transformations. Secondly, we select the salient regions in the normalized image based on feature points detected by Harris corner detector. Watermarking is performed in Discrete Cosine Transform (DCT) domain of salient regions by adapting block based embedding. The scheme is completely blind, that it does not require the original image to detect the watermark. We thus address the issue of tolerating affine transformations in medical images that is hardly be found in existing schemes of medical image watermarking. In third part of the dissertation, we investigate the complete reversible watermarking system. The proposed method uses integer wavelet transform and companding technique which can embed and recover the secret information as well as can restore the image to its pristine state. The method takes advantage of block based watermarking and iterative optimization of threshold for companding, which avoids histogram pre and post-processing. Consequently, it reduces the associated overhead usually required in most of the reversible watermarking techniques. As a result, it keeps the distortion small between the marked and the original images. Experimental results show that the proposed method outperforms the existing reversible data hiding schemes reported in the literature.