數位影像的驗證與還原

Abstract

隨著網際網路的快速發展，我們越來越容易從網路上擷取到各式各樣的影像與多媒體資訊。但是當我們取得這些影像資訊時，並無法保證這些影像的完整性；因為網際網路的安全性仍然備受爭議，我們所擷取到的影像資訊有可能是被人家竄改過的。因此，如何保護影像的完整性是一個重要的研究主題。本篇論文的主要目的就是先藉由影像驗證法來保護影像，指出影像被竄改之處，藉此達到偵測影像完整性的目的。此外，本篇論文更進一步利用資訊隱藏的技術，將被竄改的部分還原回來，達到影像還原的目的。我們共提供三個方法來達成影像驗證與還原的目的。第一個方法是利用簡單的資料隱藏法，將用來驗證與還原的資訊藏在影像的”最不重要位元”，當這些藏進去的資訊被更動時，我們就能判斷此處已被竄改，並還原此部分的資訊。第二個方法其目的則是改進第一個方法，將用來驗證與還原的資訊藏於”中等重要位元”，使客戶有較多的選擇，以增加其強固性。第三個方法則是將重要的資訊藏於頻率域中，利用離散餘弦轉換，將重要的低頻係數藏於影像中頻部分，再將影像驗證資訊藏於影像的高頻部分。最後，我們列出實驗結果來顯示所提方法的可行性。

With the rapid growth of Internet, we can get various images and multimedia data on Internet easily. When we get these images, we cannot verify the integrity of these; because the security of Internet is still a problem, and the image that we get may be tampered. Therefore, how to protect the integrity of images is an important research topic. The thesis first uses image verification to protect the images. Based on the image verification, the tampered places of the images will be pointed out. After that, the thesis can further use data hiding technology to recover the tampered place. Three methods are proposed for image verification and recovery. The first method uses a simple data hiding approach that hides the verification and recovery data into the least significant bit (LSB). When the hiding data have been changed, we can make sure that the part of the image has been tampered. The second method improves the first one by hiding the verification and recovery data into the moderately significant bit (ModSB) of the images. The two methods, combined together, give the users more choice and hence improve the robustness of the verification and recovery system. The third method hides the needed data into frequency domain. By using discrete cosine transformation (DCT), the third method hides the important low-frequency coefficients into the middle-frequency place of the image, and hides the verification data into the high frequency place of the image. Experiments show the feasibility of the proposed methods.