標題: 在影像中隱藏資料的研究
A Study on Data Hiding in Images
作者: 吳大鈞
Da-Chun Wu
蔡文祥
Wen-Hsiang Tsai
資訊科學與工程研究所
關鍵字: 資料隱藏;偽裝學;浮水印;資料嵌入;影像驗證;多基底數字轉換;人類視覺系統;Data Hiding;Steganography;Watermarking;Data embedding;Image verification;Multiple-based number conversion;Human visual system
公開日期: 1999
摘要: 隱藏資料在影像中是一種將隱密訊息或浮水印資訊嵌入數位影像中的技術。在本論文中,首先提出一個能將影像嵌入另一影像中的資料隱藏方法。再依序報告我們提出的三種可以嵌入任意型態的隱密訊息於影像中的方法。一個在影像中隱藏易碎浮水印及一個在影像中隱藏強韌浮水印的方法也包含在本論文中。這些資料隱藏技巧可用在不同的應用。 本論文中所提出的第一個方法是利用影像中相鄰像點具有相仿灰階值的特性,與人類視覺對影像的平滑區域及對比強烈區域具不同的靈敏度特質,將影像隱藏於另一影像中。這隱藏方法能無損的保藏機密影像且製造出的偽裝影像的品質退化很小。第二個方法中,以這些相同特性為基礎而發展出一個有效率的資料隱藏方法,能將一般任意型態的機密資料隱藏於影像中。機密資料被適切的嵌入於影像中未重疊的各個像素對所計算出的差值中。此方法從已嵌入隱藏資料之影像中擷取出隱藏資料時不需參考原影像。第三個方法中,另一個將資料隱藏於影像的方法被開發出。此方法是以失真性影像壓縮方法及多重基底數字轉換方法為基礎。對一影像使用任何失真性影像壓縮,產生一處理過的影像來當成資料嵌入原影像時的基準。此法提供了一個能在嵌入步驟開始進行前就能預見最遭失真情況的偽裝影像的方法,所以可以控制偽裝影像的失真程度。在第四個方法中,我們利用人類視覺模式及多重基底數字轉換方法發展出一個在影像中嵌入資料的方法。此隱藏方法保證人眼無法察覺出已嵌入資料之影像的變化。此種隱藏方法在擷取隱藏資料過程,也不需參考原影像。在我們所提出的第五個方法中,此人類視覺模式被用來發展一個易碎的數位影像浮水印技術來作影像確認及竄改偵測。此方法能偵測出任何的影像改變及指出改變的位置。最後,此視覺模式也被應用在開發一個強韌的數位浮水印方法,此方法可在影像中隱藏一個視覺無法感知的強韌浮水印。為了使以上所有提出的方法具更高的安全性,亂數技巧也可以被使用來達到傳統密碼學的效果。 實驗證明,本論文所提出的六個方法具有可行性與有效性。
Data hiding in images is a type of technique that embeds secret messages or watermark information in digital images. In this dissertation, a data hiding method that can be used to embed images into cover images is proposed first. Three approaches to embed any form of secret messages are then reported sequentially. A method for hiding fragile watermarks in images and a method for hiding robust watermarks in images are also included. These data hiding techniques are useful for different applications. The characteristic of gray value similarity in consecutive image pixels as well as the variation of the human vision's sensitivity from smooth to contrastive are exploited in the first proposed approach for hiding images into cover images. The hiding method preserves secret images with no loss and produces stego-images with low degradation. In the second approach, an efficient data hiding method that is based on the same characteristic is developed for embedding any form of secret messages into a cover image. Secret data is properly embedded in each difference value which is calculated from every non-overlapping pixel pair of the cover image. The embedded data can be extracted out from the resulting stego-image without referencing the original image. In the third approach, another method of data hiding in images is developed, which is based on lossy compression and a multiple-based number conversion method. Any lossy image compressor may be applied to a cover image to produce a result as the basis for embedding data in the cover image. The method provides a way to foresee the worst distorted stego-image before the embedding steps begin and thus control the degree of distortion in the stego-image. The fourth approach is developed by using a human visual model and a multiple-based number conversion method for embedding data in images. It is guaranteed that the changes in the stego-images are imperceptible. The embedded data can also be extracted out without referencing the original image. The human visual model is also employed in the fifth approach to develop a fragile watermarking technique for image verification and tamper detection. Any alternation of the image can be detected and localized. Finally, a robust watermarking method based on the previously-mentioned human visual model is proposed in the sixth approach. The method can be employed to create transparent robust watermarks in images. To achieve higher security in all of the proposed approaches, pseudo random mechanisms are employed to achieve cryptography. Experimental results show that the proposed approached are feasible and effective.
URI: http://140.113.39.130/cdrfb3/record/nctu/#NT880394005
http://hdl.handle.net/11536/65497
顯示於類別:畢業論文