摘要
在本論文中，我們提出一些基於頻率域之改良式數位浮水印技術。隨著網際網路的發達與普及，有愈來愈多的通訊傳輸，是經由網際網路所進行。但基於頻寬有限的考量下，數位資料在傳輸前絕大部分都會進行壓縮的動作，而其中JPEG是最普遍與最常見的檔案壓縮標準。因此，數位浮水印機制的基本功能，便是要能抵抗JPEG的壓縮攻擊，也就是強健性的要求，進而能達到數位資料之著作權的保護。然而，在JPEG量化過程中，太過相近的係數值將會被量化在同一位階上，此現象對於作用在頻率域的數位浮水印技術而言，常常會有產生誤差的問題。 另外，為符合實際的應用，數位浮水印的演算法，也大都會被要求為盲目形式，即在不需要原始影像下，就能萃取出數位浮水印的技術。除此之外，另一個重要的課題，就是如何在浮水印影像與萃取出的數位浮水印之間，讓兩者取得最佳的結果。 在本論文提出了幾種改良式數位浮水印技術，首先是利用兩個輔助的變數，針對量化後的係數值進行調整以達到誤差的改善。接著第二個方法是針對係數的調整過程中，利用絕對值的方法，防止過大的變動所產生的變號問題。最後在第三個方法中，本論文提出一種結合Reed-Solomon碼(RS碼)調變於數位浮水印的技術，利用頻率域中的直流係數、低頻係數以及區塊間的相關性來建立出RS codes的資料向量(message)，此資料向量經由RS編碼過後將產生出一碼字向量(codeword)，而冗餘資料向量 (redundancy)再依據浮水印資料來進行調整，嵌入到影像中。然後，我們以RS解碼器的結果，做為判斷浮水印的依據。在介紹每一個方法時，均提供模擬的實驗結果，以說明本論文提出的方法之可行性。
並列摘要
In this dissertation, several improved and advanced methods for the watermarking techniques are presented. With the fast development and popularity of internet, more and more information communication and transmission are carried out by internet. It is well known that due to the consideration of bandwidth, the digital signal must be compressed firstly before the signal is transmitted. In all of the methods of signal compression, JPEG (Joint Photographic Experts Group) is the most common method used for the digital image’s compression, particularly for those images produced by digital photography. JEPG could also be treated as a standard of image compression. However, the basic requirement for the robustness of the watermarking mechanism should have the ability to resist the attack of JPEG and still be able to provide the protection of digital information copyright. In the process of JPEG quantification, two similar DCT (Discrete Cosine Transformation) coefficients could possibly be quantified to the same level. To the watermarking technique on the frequency domain, such condition might cause the mistake of the extracting watermark from the lossy compression of the watermarked image. For the practical application, the watermarking technique should have the ability to extract the exact watermark without the information of the original host image. In this research, an important issue was discussed and studied. This issue is about how to make the qualities of the watermarked image and the retrieve watermark can reach the best trade-off. In this research, several methods are presented for improving the technique of watermark. First method is to adjust the quantized coefficients by using two auxiliary variables and then reach the improvement of the error. Second method uses the absolute value idea to prevent the sign change due to the over adjustment of the quantized coefficients. At last method, a watermarking technique by the modulation method based on Reed-Solomon code is proposed. The DC (direct current) value, low-frequency coefficient, the inter-block correlation are used to create the message of RS codes and then RS codeword could be produced. The redundancy vector will also be embedded into the image. The results of RS decoder will be used to judge the watermark. In the whole study, each method will be experimented and the experimental results will be used evidence the feasibility of the method proposed.
參考文獻
延伸閱讀
- 許柏偉（2006）。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〕。華藝線上圖書館。https://doi.org/10.6827/NFU.2006.00008
- 吳志堅、李勝義（2013）。可抗幾何攻擊之盲目數位浮水印技術探討。前瞻科技與管理，3(2)，101-118。https://doi.org/10.6193/JATM.2013.0302.07
- 吳正綱（2008）。數位浮水印之研究〔碩士論文，國立虎尾科技大學〕。華藝線上圖書館。https://www.airitilibrary.com/Article/Detail?DocID=U0028-1107200811423200
- 葉信呈（2016）。基於差值擴張方法之可視可逆浮水印技術研究〔碩士論文，淡江大學〕。華藝線上圖書館。https://www.airitilibrary.com/Article/Detail?DocID=U0002-0109201623442600
- Yen, S. H., & Wang, C. J. (2006). SVM Based Watermarking Technique. 淡江理工學刊, 9(2), 141-150. https://doi.org/10.6180/jase.2006.9.2.08