利用可復原浮水印技術於醫學影像竄改之偵測與修補

Tamper Detection and Inpainting for Medical Images Using Recoverable Watermarking Techniques

葉紹康(Shao-Kang Yeh)

指導教授 ： 繆紹綱

中原大學/工學院/電子工程研究所/碩士(2009年)

https://doi.org/10.6840/CYCU.2009.00824

摘要

現代的電子病歷包含了病人的資訊和醫療歷史紀錄，包括病人的姓名、醫師的處方、病例史、檢查的報告、治療的程序、診斷的影像等等。其中診斷影像也稱作醫學影像，它可輕易地藉由使用一般的影像處理軟體在電腦上做修改。醫院、保險公司，以及病人可能因為各種理由去修改，而被篡改的影像也可能違法被使用，造成個人或機構的重大損失。更糟的是，這些醫學影像很容易被惡意人士非法的修改而且不被察覺。因此，如何有效偵測影像被竄改是個值得探討的重要問題。
針對此問題，本論文提出同時結合脆弱型以及強健型浮水印的方法，並依照醫學影像之區域重要性區分感興趣區域(ROI)及非感興趣區域(NROI)。其中脆弱型的浮水印嵌在ROI，同時將偵測竄改與竄改區域復原的資訊也嵌入ROI，用來做資料完整性的確認，例如偵測局部修改，且不需要和原始影像比較。在取出ROI部分的浮水印之後可以接近無失真的方式復原回原始值；強健型的浮水印嵌入NROI，對於醫生診斷並無影響。NROI使用強健性的嵌入方法對一些影像處理或影像壓縮的攻擊有極高的強韌性，還原之後對原始影像的視覺品質也不會造成任何破壞。
實驗結果顯示所提方法能夠偵測出醫學影像中局部的修改，且成功的以接近無失真的方式復原ROI部分的原始值，並在系統偵測出竄改時可修復竄改區域，對於散佈的醫學影像資訊能夠增加它的安全性和完整性。

關鍵字

感興趣區域 ； 數位浮水印 ； 醫學影像

並列摘要

A modern electronic patient record contains personal information and health history of a patient, including name of the patient, physical examinations, prescriptions, historic pathology, laboratory examination reports, treatment procedure, diagnostic images, and so on. Diagnostic images, also called medical images, can be modified very easily by using an ordinary image processing software in a computer. Hospitals, insurance companies, as well as patients might want to modify medical images for various reasons. The tampered images may be used illegally and cause significant losses and troubles for an individual or organization. In the worst case, malicious people can easily modify these medical images illegally without being discovered. Therefore, how to detect a tampered image effectively is an important problem that deserves serious investigation.
For this problem, this study attempts to propose a solution that combines both the fragile and robust watermarking techniques, and each medical image is divided into region of interest (ROI) and region of no interest (NROI) according to their diagnostic importance. A fragile watermark is embedded in ROI, and the tamper detection and recovery information is also embedded in ROI. This information is used for data integrity verification, such as the detection of local modification, without having to compare with an original image. A watermarked image could be converted into its corresponding original image near losslessly after the extraction of embedded data in ROI. A robust watermark is embedded in NROI, which has no impact on a doctor’s diagnosis. The proposed robust embeding method is robust enough to resist image processing and compression attacks, and the original image has no visual quality degradation after recovery.
Experimental results show that the proposed method can detect the tampered locations of the medical image, and successfully recover the original image pixels with almost no distortion in ROI. The proposed system can fix tampered portions in the image after tamper detection, and provide enhanced security and integrity for the distribution of medical image data.

並列關鍵字

region of interest ； medical images ； digital watermark

參考文獻

[1] S. Kaisara, “Realization of the Computerized Patient Record: Relevance and unsolved Problems,” Int. J. Medical Informatics, Issue 1, vol. 49, pp. 1-8, Mar. 1998.

[7] R. C. Gonzalez and R. E. Woods, Digital Image Processing, 2nd ed., Prentice-Hall, 2002.

[10] H. M. Chang and L. H. Chen, “An Invisible Watermarking Method for Color Image,” Proc. of IPPR Conf. on Computer Vision, Graphics, and Image Processing, pp. 22-29, 1999.

[11] C. F. Wu and W. S. Hsieh, “Digital Watermarking Using Zerotree of DCT,” IEEE Trans. on Consumer Electronics, vol. 46, no. 1, pp. 87-94, 2000.

[12] D. C. Lou, J. L. Liu, and M. C. Chang, “Digital Watermarking Using Multi-Resolution Wavelet Transform,” Proc. of IEEE 37th Annual Int. Carnahan Conf. on Security Technology, pp. 370-377, Oct. 2003.