طراحی یک روش الگوگذاری دیجیتال خود ترمیم جهت احراز اصالت اسناد چاپی و دستنویس الکترونیکی

هدف: پژوهش حاضر با هدف بررسی فعالیت ‌های انجام شده در خصوص الگوگذاری دیجیتال اسناد و متون به منظور اثبات اصالت اسناد و حفاظت حق نشر و ارائه یک مدل جدید الگوگذاری دیجیتال متون فارسی چاپی و دستنویس با قابلیت تشخیص جعل سند و اصلاح تغییرات انجام شد. روش شناسی: این پژوهش از نوع کاربردی است. برای شبیه ‌سازی الگوریتم از 20 متن دستنویس فارسی و 35 متن چاپی استفاده شد. در الگوریتم پیشنهادی با استفاده از اطلاعات نقاط تصویر و ترکیب آن با یک کلید مخفی، الگوی دیجیتال تولید و در تصویر متنی ذخیره شد. سپس تصویر مورد حملات تعمدی قرار گرفته و الگوی پنهان شده در متن بازیابی گردید. در نهایت با توجه به قابلیت خود ترمیمی روش، نواحی جعل شده تصحیح و سند به شکل اولیه بازگردانده شد. یافته‌ ها: دقت روش پیشنهادی در تشخیص و تصحیح نواحی جعل شده بر حسب دو معیار دقت تشخیص (tda) و دقت ترمیم (tra) محاسبه شد. الگوریتم در حالت میانگین قادر به شناسایی نقاط دستکاری شده با دقت 93.82 درصد و تصحیح و بازیابی تغییرات با دقت 92.27 درصد بود. نامحسوس بودن الگو بر حسب مقیاس psnr برابر با 40.18 دسی بل به دست آمد. نتیجه ‌گیری: نتایج حاصل از اندازه گیری عددی دقت مدل پیشنهادی و همچنین نتایج شهودی حاکی از کارایی مناسب آن است. الگوی ادغامی به صورت نامحسوس و با حفظ کیفیت بصری اسناد درج می ‌گردد و برای تمامی متون دستنویس و چاپی الکترونیکی و برای کتابخانه ‌های دیجیتال، آموزش الکترونیک و دولت الکترونیک قابل پیاده‌‌سازی و اجرا می‌ باشد.

A Selfrecovery Digital Watermarking Approach for Tamper Detection of Handwritten and Printed Electronic Documents

Objective: The aim of this study is to present a brief discussion of the digital image watermarking for texts in order to prove the documents authentication and copyright protection and to introduce a new method of digital watermarking of printed and handwritten Persian documents image for tamper detection and recovery. Methodology: This paper is based on an applied research.In this study, 20 Persian manuscripts and 35 printed texts were used to simulate the algorithm. In the proposed algorithm, using the information of the image pixels and combining it with a hidden key, we produced the digital watermark and embedded it in the text image. We then intentionally attacked the image and extracted the hidden watermark in the text. Finally, due to the selfrecovery capability of the method, the tampered areas were detected and the document was restored to its original form. Findings: In all 55 test images, separate watermark were embedded according to the proposed algorithm. To evaluate proposed system the three types of attacks includes adding, removing and updating, 30 random images were selected. In 20 images, all three types of manipulation were applied. Then the watermark was extracted and the results were studied. The accuracy of proposed method in detecting and correcting tampered areas was calculated according to two metrics: Tamper Detection Accuracy (TDA) and Tamper Recognition Accuracy (TRA). The algorithm was able to identify manipulated areas with an accuracy of 93.82 percent and correct and recover changes with an accuracy of 92.27 percent. It is important to note that the results obtained are based on the number of correctly identified bits. Therefore, the accuracy of less than 100 percent does not mean inability to detect some manipulated areas; Rather, as can be deduced from the intuitive results, this means that despite the correct diagnosis of the tampered area, only some pixels of that area have not been correctly detected or recovered and therefore do not have a negative effect on the final tamper proofing. For imperceptibility test, PSNR value is estimated. The average value of PSNR in the suggested method is 40.18 dB. Conclusion: Determining the documents authenticity without the availability of the original documents is one of the main necessities. Digital watermarking is one of the most powerful tools for document authentication and protection of copyright. Our proposed method has two special features that are considered to be its advantages over other methods. The first feature is the ability to work with all kinds of Persian text images. In addition to the ability to run on printed texts, it is also possible to model in handwritten texts as well as the combination of text and image. Another advantage is good rate of tamper detection and the ability to selfrecovery of tampered images. The numerical and intuitive results indicate the efficiency of proposed scheme and so can be implemented for all handwritten and electronic printed texts. Therefore, this method can be used to protect the security of printed and handwritten documents in digital libraries, elearning, egovernment and any electronic document exchange system.