利用影像不變特徵與區域相關性之數位影像鑑識方法

Digital Image Forensic Method Using Image Invariant Feature and Region Correlation

廖展章(Jhan-Jhang Liao)

指導教授 ： 郭天穎

國立臺北科技大學/電資學院/電機工程系研究所/碩士(2012年)

摘要

數位化時代的來臨造就多采多姿的數位生活，但由於數位影像易於修改，因此數位影像的真實性成為重要的議題。區域複製篡改(Region Duplication Forgery)為一種簡單又常見的影像篡改方式，近年來提出的影像區域複製篡改偵測方法，大多使用稀疏性影像不變特徵，如SIFT、SURF等作為偵測之依據，雖然這些方法能有效偵測帶有幾何與亮度改變等篡改操作，但在影像不變特徵數量過於稀少或影像具有相似重複物件(Intrinsic Repeated Elements)時，容易造成無法偵測或錯誤判斷的結果。
本論文藉由修改SIFT演算法及提出影像不變特徵分群分析(Cluster Analysis)與區域離群值檢測(Local Outlier Detection)，以改善現有文獻無法解決之問題。此外，我們基於影像區域相關性與影像紋理特徵設計一適應性的篡改區域定位方法，用以獲得最佳的篡改區域定位圖作為偵測之結果。在實驗測試中，我們設計帶有幾何轉換與亮度變化的自動篡改程式進行實驗評估，從實驗結果顯示，我們提出利用影像不變特徵與區域相關性之偵測方式較現有文獻方法更加強健與準確。

關鍵字

影像篡改偵測 ； 影像不變特徵 ； 尺度不變特徵轉換(SIFT) ； 分群分析 ； 區域離群值檢測

並列摘要

With the coming of age of the digital era, it brings us a colorful digital life. However, the authenticity of digital image has become an important issue as the digital image is easily modified. Region duplication is a common and simple way to modify or tamper the digital image. The recent methods proposed in literature are based on spare image invariant feature (such as SIFT, SURF), and can effectively detect the geometric and brightness tampering. However, they fail to detect the tampering when the image invariant feature is inadequate, and often misclassify the original contents as the duplication tampering when the image contains the intrinsic repeated elements.
Our method proposes a modified SIFT algorithm, an image invariant feature clustering analysis, and local outlier detection to improve the above problem. In order to locate the tampering region, we design an adaptive tampering locating method based on image local region correlation and image texture feature. We evaluate our proposed approach on tampered images with and without intrinsic repeated elements, and the geometry and brightness of the tampered duplicated region is further altered by an automatic forgery program. The experimental results and analyses demonstrate that our proposed method is robust and effective in region duplication detection.

並列關鍵字

Image Tampering Detection ； Image Invariant Feature ； Scale-Invariant Feature Transform (SIFT) ； Clustering Analysis ； Local Outlier Detection ； Image Texture Feature

參考文獻

[1] A. Haouzia and R. Noumeir, “Methods for Image Authentication: A Survey,” Springer Netherlands, Multimedia Tools and Applications, Vol. 39. No. 1, pp. 1-46, Aug. 2008.

[2] H. Farid, “Digital Image Forensics,” Scientific American Magazine, pp. 66-71, Jun. 2008.

[3] H. T. Sencar and N. Memon, “Overview of State-of-the-Art in Digital Image Forensics,” Statistical Science and Interdisciplinary Research. World Scientific Press, Singapore, 2008.

[4] B. Mahdian and S. Saic, “Blind Methods for Detecting Image Fakery,” IEEE International Carnahan Conference on Security Technology, pp. 280-286, Oct. 2008.

[5] T. V. Lanh, K. S. Chong, S. Emmanuel and M. S. Kankanhalli, “A Survey on Digital Camera Image Forensic Methods,” IEEE International Conference on Multimedia and Expo, pp. 16-19, Jul. 2007.