| 摘要: | 視覺密碼是一種既安全又獨特的秘密保護與分享方法，有別於傳統的密碼學方法，視覺密碼的解密過程只需利用人眼就可以解密，而不需要任何複雜的解密演算法或是電腦設備的協助。因此，在無法取得電腦設備的時候，視覺密碼的方法就非常具有實用性。 在視覺密碼的研究中，像素擴展與對比是兩個非常重要的研究主題。雖然像素擴展是建構視覺密碼機制的重要技術，但是它卻會導致影像變形、浪費儲存空間與難以攜帶等問題。近年來已經有一些不需要像素擴展的方法相繼被提出，其中有許多方法直接利用現成的機制，並配合機率的觀念來達成影像不擴展的目的，因此這些方法所能達到的影像對比，與現成的機制所能達到的對比是完全一樣的。雖然有些視覺密碼方法在某些秘密分享結構上能夠提供最佳對比，然而還是有許多方法並不保證具有最佳的對比，因此，本研究將針對像素擴展與對比的問題提出一個解決的方法。我們將針對任意的秘密分享結構，以機率的觀念來建構多目標線性規劃模型，並且利用目標規劃的方法來求解最佳對比。相較於Ateniese et al.的方法，我們的方法不但可以避免影像擴展，而且就平均而言，還能獲得更好的對比。 由於視覺密碼方法會使重建後的秘密影像產生對比損失，因此，對於灰階影像而言，如果其灰階值範圍太過於狹窄的話，將會使得重建後的影像更加難以被人眼辨識。因此，我們在本研究中也將提出一個門檻式區塊加密程序，以解決灰階影像對比損失的問題。 近年來，視覺密碼及其觀念已經被大量用來保護數位影像的智慧財產權，然而，某些以視覺密碼為基礎的方法並無法滿足視覺密碼的安全條件，因此這些方法其實是無法被用來保護數位影像的智慧財產權。此外，大部分的方法都只適合用來處理黑白或是少數幾種顏色的浮水印，而不適合用來處理灰階或是彩色的浮水印。因此，本研究將利用樣本均數抽樣分配的相關理論與特性，來解決安全性與彩色浮水印的問題。在我們的方法中，樣本均數抽樣分配的相關理論與特性，將被用來滿足安全性與強韌性的要求。基本上，我們的方法並不會真的將浮水印藏入影像中，而是由受保護的影像中以抽樣的方式來產生所有權影像，以做為將來驗證影像所有權的依據，在驗證影像所有權時，我們的方法並不需要原圖的協助。由於我們的方法不會改變受保護影像的內容，因此我們可以為單一張影像註冊多個所有權，而不會破壞其他的浮水印。這個特性使得我們的方法非常適合應用在那些不允許修改受保護影像內容的場合。最後，實驗結果將顯示我們的方法可以抵抗許多種常見的影像處理運算或是攻擊。 Visual cryptography is a very secure and unique way to protect secrets. Unlike traditional cryptographic schemes, it uses human eyes to recover the secret without any complex decryption algorithms and the aid of computers. Thus, when computers or any other decryption devices are not available, visual cryptography schemes can be very useful. In the study of visual cryptography, pixel expansion and contrast are two primary issues. The technique of pixel expansion has been largely adopted to construct visual cryptography schemes. However, pixel expansion can result in many problems such as the problem of image distortion, the requirement of more storage space, and the difficulty in carrying shares. Thus, some methods were proposed to cope with the problems of pixel expansion. Many of these methods integrate the probability concept with the ready-made schemes to avoid pixel expansion. In such methods, the contrast is the same with that provided by the underlying schemes. Although some schemes can provide the best contrast for some access structures, there are still many other schemes can not provide the best contrast. Therefore, in this dissertation, a multi-objective linear programming model for general access structures is constructed to deal with the problems of pixel expansion and contrast. Then, the solution space is searched by the modified simplex method, which is a kind of goal programming algorithms. Experimental results will show that, in average, the proposed method has better contrast than Ateniese et al.’s. It should be noted that the method of visual cryptography would result in contrast loss of the recovered secret images. If the secret image to be encrypted is a gray-level image with a narrow dynamic range in its gray scales, the phenomenon of contrast loss can be a serious problem because the recovered secret image may be difficult to be identified. Therefore, in this dissertation, a block encryption procedure with thresholding techniques is proposed to compensate the losing contrast for gray-level images. Recently, visual cryptography or its concept is largely adopted to protect the intellectual property rights for digital images. However, some of the methods do not satisfy the security conditions of visual cryptography; thus they can not be used to protect the copyright of the images. Moreover, many methods are not suitable for gray-level or color watermarks. Therefore, in this dissertation, we come up with a solution to deal with the problems of security and color watermarks. In our scheme, the theory and properties of sampling distribution of means are employed to satisfy the requirements of security and robustness. The proposed method does not really embed the watermark into the host image. Instead, an ownership share is generated from the host image as a key to reveal the watermark without resorting to the original image. This property also allows multiple watermarks to be registered for a single host image without destroying other hidden watermarks. Since the host image will not be altered, the proposed scheme is very suitable for the applications where modifications to the image are not allowed. Finally, the experimental results will show that our method can resist many common attacks. |
