The Techniques of Digital Image Authentication and Watermarking

[[abstract]]為了驗證經由網路傳輸的數位影像之完整性，本學位論文提出兩個影像驗證技術，第一個方法為植基於壓縮與預測機制之影像驗證技術，主要針對影像區塊產生特殊的壓縮驗證碼，接著，將驗證碼與相關還原資訊嵌入於數位影像。當之後接收到可疑的數位影像時，可以藉由檢查驗證碼之完整性來檢驗該影像是否遭到竄改，並且利用從該影像中取出的還原資訊來大致恢復那些被竄改之區塊；第二個方法為改良且植基於數獨規則的影像驗證技術，主要根據每個影像區塊的特徵來修改區塊中的像素值，使其符合數獨之規則，當接收者收到可疑的影像時，便可以藉由數獨規則來檢查個別區塊是否遭受惡意竄改。實驗結果顯示，提出的兩個影像驗證技術皆能使得加入驗證碼後的影像仍維持不錯的品質，而且能有效地、精準地標示出影像中被竄改的位置。[[abstract]]In order to verify the integrity of digital images that are transmitted via the Internet, this thesis proposed two image authentication schemes. One of them is an image authentication technique based on a data compression and prediction mechanism. For image blocks, first is to produce special verification codes, and next, these codes and other information related to data recovery will be embedded into the image. When someone receives a suspicious image, he/she can check the integrity of embedded verification codes to judge whether the image was tampered with or not. Further, the recovery information extracted from images can be used to restore the tampered area. Another is an improved image authentication scheme based on Sudoku rules. Each image block is first performed the quad-tree analysis according to its unique characteristic, and then pixels in the block are modified to fit in with Sudoku rules. When a receiver obtains a suspicious image, he/she can check if an individual block has been tampered with by using the same Sudoku rules. Experimental results show that the proposed two schemes not only maintain good quality of the images, but also are able to point out the tampered area accurately and efficiently