基於差值擴張與二維模型於醫學影像之可逆資訊隱藏技術

Abstract

資訊隱藏技術是一種用於內容認證、保護資訊的方法，在某些應用上，此技術必須是可逆的，像是軍事或是醫療。一般來說，醫學影像中包含許多純黑跟純白的點，而這些點在資訊隱藏技術中有極大的機率會導致溢位的情形發生。 在本論文中，我們提出一個結合差值擴張法與二維差值擴張模型的可逆資訊隱藏技術，可避免溢位的發生。首先我們將影像先分成許多區塊，每個區塊為4×4像素，再根據其平均值及標準差將區塊分為5種類型，其中4種為可嵌入的，1種為不可嵌入的。另外，我們也提出了一個以數學方式證明能有效的區分此5種類型的方法。實驗結果顯示出本論文所提出的方法相對於其他方法在大多數的醫學影像上，能達到較高的嵌入容量及影像品質，且產生較少的位置圖資料量。此外，我們也將此套系統發展成在Android平台上的應用程式。Data hiding is a method used for content authentication and information protection. For some applications, such as the medical or military system, the data hiding is required to be reversible. In general, the medical image consists of many pure black and white points which leads to that a data hiding may cause the overflow and underflow problems. In this paper, we proposed a new reversible data hiding method for medical images which is based on the difference expansion (DE) method and the two-dimension difference expansion (2D-DE) scheme. Firstly, the image is divided into blocks of 4×4 pixels. Based on the average and standard deviation of each block, the blocks are categorized into five types. Four of them are embedded blocks and the remaining one is a non-embedded block. We also proposed a mathematical approach to efficiently differentiate these five types. Experimental results show that the proposed method compared to the others can achieve higher embedding capacity and generate less the size of location map for most medical images without losing the image quality. Furthermore, we develop this system as an app on Android OS.