浮水印的賽局演算與印表機來源鑑識

Abstract

數位影像處理常使用多種技術整合，且廣泛的應用於各領域中，本論文將依常見且實用的兩種數位影像處理技術進行研究與探討：以浮水印的賽局演算及印表機來源鑑識為例。 第一部份，我們考量了浮水印半透明性與強韌性兩關鍵因素之平衡，提出一個以人類視覺系統與雜訊可見函數為主的適性化可視浮水印演算法，並以賽局理論為基礎，發展一個浮水印的賽局演算架構，透過模擬嵌入浮水印之環境，分析浮水印對影像之影響以及攻擊者可能採取的攻擊行為，進而獲得最佳的浮水印嵌入能量，有效的提供浮水印嵌入者做決策。第二部份，我們結合原先已有的數位相機與印表機圖片鑑別研究成果，以中文輸出文件為例，透過不同影像處理與特徵值擷取的方法來分析數位內容，利用計算灰階共生矩陣與離散小波轉換的統計方法取得34個影像特徵值。接著，我們進一步的使用特徵值選擇法來獲取較具影響力之特徵值，最後經由支持向量機的分類技術進行印表機設備來源之鑑識。 由大量的實驗結果，證明本論文所提出的浮水印方法架構與鑑識印表機來源之有效性；所提出之賽局式浮水印架構能解決先前浮水印能量與抗攻擊能力之問題，而印表機來源鑑識方面能獲取最重要之特徵進而產生良好之辨識效果，因此本論文在數位內容所有權保護研究和數位鑑識兩方面具有重要的貢獻。

Digital image processing generally applies various technologies which are widely used in many different fields. This thesis dicusses two topics of digital image processing for research purposes which include a game-theory based watermarking system and the forensics for printer source identification. In the first part of this study, we explore the inter-relationship between the robustness and translucence of visible watermarking and propose a novel watermarking algorithm named Adaptive Content and Contrast Aware (ACOCOA), which combines Human Visual System (HVS) and Noise Visibility Function (NVF). ACOCOA’s architecture uses game theory to resolve the conflicting interest of intensity and perceptual efficiency which provides the selection of optimal watermark intensity insertion for the decision maker. The second part of this study is to identify the printer sources for Chinese characters by applying statistical analysis on the Gray Level Co-occurrence Matrix (GLCM) and Discrete Wavelet Transform (DWT), which involves image processing techniques and data exploration methods to obtain 34 image features. Subsequently, the feature selection procedures are used to get a usable subset while only the most important features are systematically retrived. Finally, the classification technique of support vector machine (SVM) is utilized to identify the printer source. The experimental results demonstrate that the feasibility of the proposed approach: the game-theory architecture for visible watermarking system not only solves the previous problem of the watermarking energy insertion but also provides effectiveness and robustness for the watermarked images under attacks. In addition, the proposed method of the Chinese printed source identification analyzes the image characteristics of the printed documents to select the most important features for gaining the high accuracy ratio of source identification. In summary, this dissertation provides substantial contribution to the research field of digital content ownership protection and digital forensics.