Studies On Watermarking Algorithm For Medical Image Content Protection

As the rapid development of medical imaging technology and the Internet, moreand more new systems appear, for example, Picture Archiving and CommunicationSystems (PACS), the medical image in the global scope capture, storage andtransmission becomes more and more convenient. This can lead to medical imagesillegal capture and patient information malicious tampering more and more easily. Sothe protection of medical image content and information will confront deepchallenges. In order to more effectively protect of medical image content andinformation, applying the watermarking technology of digital image to the field ofmedical image protection has become a research focus. The watermarking technologyof digital image is an effective method that embeds specific values as the copyrightauthentication information into the digital images. It realizes hiding informationwithout affecting the image quality and provides the authentication basis for copyright.Medical image is a special kind of digital image, therefore, the watermarkingtechnology of digital image can provide technical support for the protection ofmedical image content and information. In the paper, based on research on thewatermarking technology of digital image and summarize the existing medical imagewatermarking algorithm, then propose the following medical image watermarkalgorithms by combining with characteristics of medical image.1. Based on the analysis that medical images of the human organ content hasimportant and non-important regions, a multiple watermark scheme for medicalimages is proposed based on DWT and DCT. The medical image is divided into theregion of interest (ROI) and region of non-interest (RONI). Embed imagecharacteristic into the ROI using semi-fragile watermarking algorithm to realize thedetection and location of image tampering, embed the patient’s information intoRONI using robust watermarking algorithm to enhance the ability against commonattacks. Firstly, a2-level discrete wavelet transform (DWT) is applied in the ROI,then the watermark is embedded using the low-frequency sub-band (LL2) of the DWTthrough quantizing the low-frequency coefficients. Finally, the RONI is divided intonon-overlapping8×8blocks. A discrete cosine transform (DCT) is used to eachblock, and the DCT coefficients of each block are sorted using zigzag scanning. Sevenintermediate frequency coefficients are selected to embed the watermark by quantizing them. Experimental results showed that our scheme has the advantages oflower error rate, strong robustness and high capacity compared with other algorithms.2. Zero watermarking is a new digital watermarking technology withoutmodifying the data of original image. For data of medical image can’t be changed, wepropose a medical image zero-watermarking algorithm based on invariant moments.First, two kinds of invariant moments are called Krawtchouk moments and Humoments are proposed. Each of them has different advantages under different attacks,so we combine stable invariants into new invariants. Then, we select a threshold toquantify the values of moment invariants and convert them into a binary sequence.Finally, do XOR operation with the sequence and the binary watermark image toembed the watermark. Experiments show that the proposed algorithm can robustlyresist image attacks such as rotation, scale, translation and noises.3. Due to the medical image format is different from the common image formats,typically a browser can’t display the whole medical image. So it is not convenient todo the common processing of medical image. Therefore, we designed a medicalimage watermarking system that can convert medical image format into commonimage format. Because the particularity that the medical image standard DICOM hastags, we design a medical image information tracking system for telemedicine. Whenusers download the medical image in the system, the user’s name will be embeddedinto the image as a new registering tag automatically. So the medical imagesdownloading from the system are marked. If the medical image is tamperedmaliciously or theft, we can find the person who will be accountable by extracting thenew tag and other relevant information registered in the database.