藉由分享而得的影像還原技術

Abstract

數位影像在WWW的環境中可能會被攻擊者竄改資料，或者在傳輸之中發生傳輸錯誤的情形。在本篇論文中，我們提出三種基於「資訊分享」技術的影像還原方法。在第一個方法之中，我們對每個8x8影像區塊進行編碼；其後，只要未被破壞的影像區塊數目大於ㄧ門檻值，我們就可以用它們來把其他已被破壞的影像區塊給還原回來。由於第一種方法若整張影像都被拿掉，則無還原機會，所以我們又提出第二種方法「多張影像交叉還原」技術。對於n張編碼後的影像，只要有t 張影像是未被破壞的，我們就可以利用這t張影像來還原其他(n-t)張已被破壞的影像。第三種方法是第二種方法的變形；由於管理者可能希望影像具備自我修復的能力，但又不希望降低影像品質，所以我們的第三種方法以增加儲存空間的方式，對於每張輸入的影像會附加一塊類似雜訊的分存影像，而其功能亦可達到和第二種方法相同的效果。 除了上述三種方法外，我們也觀察到視覺密碼學(Visual Cryptography)和利用多項式的影像分享(Image Sharing by Polynomials)各有其優劣點；所以本論文最後提出一個結合此二者的新的分享機制；其分存影像可以依據環境狀況來採用其中一種回復方法。

In WWW environment, digital images may be tampered by attackers, or get lost during transmitting. In this thesis, we propose three kinds of image recovery methods, based on sharing techniques. In the first method, the given image is partitioned into 8-by-8 blocks, and we encode each 8-by-8 image block. Later, if the number of the non-attacked blocks is greater than a threshold value, then we can recover all the tampered blocks by using them. Because the first method cannot recover the image if it is deleted completely, we propose the second method which is the cross recovery of multiple images. For n encoded images, if t out of the n images are not-attacked, we can recover the other (n-t) images by the t non-attacked images. The second method is not error-free when each image is embedded with recovery information. The third method is a modification of the second method; and it is for the users who require the images to have cross recovery ability without degrading the images’ qualities. This method appends a noisy shadow image to each input image. Besides the three methods described above, it is also observed that Visual Cryptography and polynomial-style image sharing both have their own advantages and disadvantages. So, this thesis also proposes a novel sharing mechanism combining the two methods. The recovery can be according to the computer resources available at the scene.