Reversible Color Image and Video Data Hiding Using Difference Expansion Methods

近年來，由於科技的快速發展與網際網路的普及使得多媒體資料的取得及下載變得十分方便；然而，科技的進步雖然增進了工作效率卻也使得盜版與不法使用變得更加容易。為此，保護智慧財產權這項議題變得相當重要，換句話說，資訊隱藏技術的研究與發展是不容忽視的重要趨勢。 資訊隱藏 (Information Hiding) 是一門古老的科學，其中包含了兩大議題﹕資料隱藏 (Data Hiding) 和浮水印 (Watermarking)。資料隱藏是將重要的資料嵌入於不受人注意的寄主媒介中，藉此分散敵人的注意力﹔然而，浮水印則是將產品資訊(例如商標)隱藏在數位化的產品中以達到著作權宣告的目的。 此論文中，我們將會敘述現有的兩個資料隱藏技術，分別是由Adnan M. Alattar 和 Vasiliy Sachnev 所提出。其中一個方法是針對Jun Tian 提出的差值擴張技術(difference expansion technique)的延伸，將三個像素當作一個單位進行資料嵌入，提供了我們將此技術應用到彩色影像的靈感。另一方法則是利用四周圍的像素關係先做一個預測後再加以藏入資料，其主要的特點為藉由調整參數達到省略隱藏資料的位置標記(location map)，換言之，增加藏入的資料量(embedding capacity)。我們將針對這兩種演算法做一個比較並試圖加以改良，期許能得到更佳的表現與突破。同時，我們也希望這個新提出的方法不僅能適用於靜止的彩色影像亦能和現有的視訊技術做結合、擴大其應用層面。The booming of powerful sharing/transmission tools of digital media contents has created an urgent need to explore the algorithms of information hiding which is widely adopted in copyright protection, content authentication, and media asset management. Digital watermarking or data hiding is the process of embedding information into a digital medium such as an image, audio or video. However, for many data hiding scheme, there exist a drawback that the process usually introduces irreversible degradation of the original content. Although this degradation is slight, it may not acceptable in applications dictating high fidelity, e.g. military and medical uses, and hence, there is an essential for a reversible data embedding. In this thesis, we would present a lossless data hiding algorithm for color images and videos based on Jun Tian’s difference expansion (DE) technique. Tian used a difference expansion transform of a pair of pixels to embed a large amount of data into gray scale images. In this work, we would apply the prediction errors to a cover media with the concept of difference expansion of triplets, which is a derivation of DE method. The performance of the proposed reversible data hiding scheme is evaluated by using different images and compared with two methods, proposed by Adnan M. Alattar and V. Sachnev et al. Further, we would like to employ this algorithm to the blocked-based DCT coefficients, that is to say, the video contents. We expected the results indicate that the proposed method allowed a high embedding capacity and less image degradatio