半盲目型浮水印技術之設計

Design of Semi-blind Watermarking

楊宗家(Tsung-Chia Yang)

指導教授 ： 李金鳳

朝陽科技大學/資訊學院/資訊管理系/碩士(2009年)

摘要

近年來網際網路發展迅速，讓使用者可以透過網路發表個人創作。但由於開放式的網際網路容易遭受有心人士下載、篡改、亦或是未經原創者同意加以轉載，而這些行為已經觸犯智慧財產權。為了遏止這類非法的行為，目前有許多學者發展浮水印系統以保護數位產權。
現今學者提出的浮水印架構可分為空間域及頻率域兩大領域。以往的空間域浮水印的應用發展及演算法較為簡單，但能抵抗攻擊的強韌性較低。另一方面頻率域的浮水印技術的演算法比空間域演算較為複雜，相較起來頻率域的浮水印擁有比較好的強韌性比較能抵抗影像處理攻擊的破壞。但目前強韌性的浮水印架構增加對抗影像處理攻擊的強韌性，大部份都是需要原始影像及浮水印的輔助，因此浪費了大量的頻寬及記憶體空間，所以較不太符合實際應用面。
本研究提出兩個半盲目浮水印技術，都是需要少數的原始資訊來做為輔助浮水印的偵測。第一個提出在頻率域中利用離散小波轉換(Discrete Cosine Transformation, DCT)以關聯規則(association rules)結合掩蔽區塊(masking)進行自我嵌入(self-embedding)的半盲目型浮水印架構，此方法仍然需要浮水印的輔助，最後根據實驗結果顯示這個半盲目型浮水印的架構在面對影像裁切的影像處理攻擊時，其浮水印的相似值較低，所以本文再以進行三次的自我嵌入，並利用表決判定s(voting)來提升浮水印判斷的正確性，最後根據實驗結果顯示，確實可有效改善誤判率情形。
雖然本研究的半盲目型浮水印技術不需要原始影像的輔助，也擁有一定程度的強韌性，但仍然需要浮水印的輔助，因此本文第二個提出以離散小波轉換(Discrete Wavelet Transformation, DWT)所產生的參照影像(reference image)，在空間域上配合著正好察覺扭曲度(Just Noticeable Distortion, JND)來進行浮水嵌入，根據實驗結果顯示確實可省去浮水印的輔助的可行性僅需要少量的位置索引(location map)做為輔助，且在面對影像處理攻擊時擁有更好的強韌性。

關鍵字

自我參照 ； 正好察覺扭曲度 ； 自我嵌入 ； 掩蔽區塊 ； 半盲目

並列摘要

In recent years, the advance of computer technology and rapid development of Internet have facilitated the rapid growth of E-commerce. The Internet become more convenient, so more and more people publish their creation on the Internet. However, it is easy to be attacked and downloaded or transmitted by those who mean to do it due to the open environment network, and these illegal acts have broken the law of the Intellectual Property Rights. In order to avoid these illegal acts, there are more and more scholars involving themselves in watermark techniques.
Now, the watermarking technique has two parts, which are space domain and frequency domain. The algorithm in the space domain, formerly are easier to be developed and the algorithm is easy too, but these methods are non-robust with attack. On the other hand, the frequency domain is robust than the space domain and the algorithm is more complex. Generally the robust watermark technique needs original data to assist watermark detection; therefore it wastes lots of memories and bandwidth, so it’s not conform actual applications.
In this thesis, we proposed two semi-blind watermark techniques; these schemes need few original data to assist watermark detection. The first semi-blind watermark scheme is designed on frequency domain by Discrete Cosine Transformation (DCT) with association rules and masking block to assist self-embedding. This scheme needs the original watermark to assist the watermark detection. According to the experimental results, this semi-blind watermark scheme had confronted the image process as image cut or Gaussian noise, the similarity value was become lower. After self-embedding watermark bits three times, the voting approach can promote the watermark detection accuracy. According to the experimental results, this scheme by using self-embedding three times with voting can exactly reduce the false-positive errors.
The second proposed watermark scheme ensures no original image is required to help watermark detection. A reference image is obtained by Discrete Wavelet Transformation (DWT) first. We used the reference image with Just Noticeable Distortion (JND) to do watermark embedding in space domain. Though the semi-blind watermark scheme still needs few location maps to assist for watermark detection, the overhead is quiet small. According to the experimental results, the watermark scheme is more robust and can confront the image process as image cut or Gaussian noise or JPEG.

並列關鍵字

self-reference ； JND ； masking block ； semi-blind ； self-embedding

參考文獻

[3] 許柏偉，植基於關聯規則之數位浮水印技術，碩士論文，國立虎尾科技大學資訊管理系，雲林，2006。

[4] 潘正祥、張真誠、林詠章，挑戰影像處理數位浮水印技術，美商麥格羅．希爾國際股份有限公司，2007。

[5] R. Agrawal, T. Imielinski, and A. Swami (1993), “Mining association rules between sets of items in large databases,” Proceeding of the ACM SIGMOD Conference on Management of Data, pp. 207-216.

[6] S. Agreste and G. Andaloro (2008), “A new approach to pre-processing digital image for wavelet-based watermark,” Journal of Computational and Applied Mathematics, Vol. 221, pp. 274-283.

[7] M. Barni, F. Bartolini, V. Cappellini, and A. Piva (1998), “A DCT-domain system for robust image watermarking,” Signal Processing, Vol. 66, pp. 257-372.