A Dual Watermarking Approach for DICOM Images

In this paper, we come up with a new watermarking method combining a zero-watermarking approach with a non-zero-watermarking approach for the purpose of authentication and identification of medical DICOM images, as well as the confidentiality of patient information. Indeed, the storage and transmission of medical images need strong confidentiality, authentication and integrity. In the proposed approach, pertinent features extracted from the DICOM image are used, on the one hand, for the zero-watermarking based on the Jacobian model, and on the other hand, to construct a watermark embedded in the black background region using linear interpolation technique. The watermark is encapsulated only in the black background region in order to avoid affecting the anatomical part whose modification may cause a wrong diagnosis. The main contribution of this paper is the method used to generate a robust watermark based on the Jacobian model and pertinent features extracted from the medical image. Experimental results reveal the quality of the method proposed: high PSNR (peak signal to noise ratio) and SSIM (similarity structure index measure) values as well as high robustness against different attacks.

I. Introduction

Nowadays, transmission of digital image information in hospitals becomes more and more developed by using modern healthcare electronic systems which facilitate acquisition, storage and transmission of medical images and information to doctors and radiologist for diagnosis and treatment. Medical image comprises two regions: the region of interest (ROI) that contains the information related to patient’s diagnosis and the region of non-interest (RONI), that does not contain important information for diagnosis. The advancement in information and communication technologies allows easy access, manipulation and replication of medical images. So, these images, especially their ROI part, need to be secured to avoid wrong diagnosis. Digital watermarking embeds additional data called watermark to protect the ownership of the image and verify its authenticity and integrity. Medical image watermarking should be used carefully, as modifying the Region of Interest (ROI) may lead to misdiagnosis [2]. Furthermore, the embedding of the watermark should be invisible. The watermark must be robust against both intentional and non-intentional attacks. Medical image watermarking can be classified into four main categories [2]: watermarking in RONI [3], [4], [5], reversible watermarking [7], [8], [9], [10], zero-watermarking [6], [7], [8], [9], [10], [11] and watermarking with image change not affecting diagnosis [13]. In this paper, we propose a watermarking method combining a zero-watermarking method with a non-zero watermarking method. The zero-watermarking is applied to the anatomical part of the image whereas the non-zero watermarking is applied to the rest of the image. By securing the anatomical part with a zero-watermarking method, we ensure that this part is not modified. Indeed, zero-watermarking does not embed the watermark directly into the original image [3]. The organization of the paper is the following: Section 2 describes related works. Section 3 presents the proposed watermarking method. Section 4 exposes experimental results. The paper ends by a conclusion in Section 5.