本論文著重在複製移動竄改偵測技術之參數探討，本論文方法將實驗影像使用SIFT計算出關鍵點，找到的關鍵點根據強度大小進行分組，分組後的關鍵點計算動量值進行比較，如通過門檻值則進行區塊擴張運算，最後標記偵測出的複製區塊。
    經過多次實驗，本研究發現在不同的實驗影像設定相異的參數以及門檻值能呈現較好的偵測結果，且在討論及蒐集資料時，發現多種因素都會影響偵測結果，大致可以分成三部分，第一部分為動量值的選取，本論文進行比對的是Hu不變動量值及Zernike不變動量值；第二部分為關鍵點，其中包含找到更少更準確的關鍵點以及根據關鍵點強度做分組及配對；第三部分為區塊擴張，其中包含進行區塊比對時，起始及展開的區塊大小選擇。
    本論文總結上述實驗，提出最佳參數的選擇方式，用以獲得最佳的複製區域偵測效果。

The thesis investigates features on detecting copy-move duplicated regions. The structure of copy-move detection is searching keypoints through the Scale Invariant Feature Transform(SIFT), matched blocks acquired from these keypoints by invariant moments, region growing by surrounding matched blocks. The analyses include the Scale Invariant Feature Transform(SIFT) for calculating keypoints, keypoints match, invariant moments comparisons, sizes of region growing blocks. This thesis examines various parameters and thresholds of the adopted structure. We also find that there are many factors to affect the detected results. Three conclusions are summarized. First, Hu’s invariant moments are better than Zernike invariant moments. Second, positions of duplicated regions can be acquired from keypoints through a robust neighboring search. Third, the optimal growing block size is then acquired. At last, a set of optimal parameters are found through the exhausted experimental results.

目錄
目錄	III
圖目錄	III
表目錄	VI
第一章	緒論	1
1.1	研究背景與目的	1
1.2	論文架構	3
第二章	相關研究	4
2.1	Hu不變動量值[2]	4
2.2	Zernike不變動量值[13]	6
2.3	Chen等人所提出的方法[3]	9
2.4	尺度不變特徵轉換[4]	9
第三章	本論文提出方法之分析	18
3.1	以多重區塊特徵進行複製移動竄改偵測技術探討	18
3.2	參數分析	18
第四章	實驗結果與比較	20
4.1	不變動量值的探討	20
4.2	關鍵點的探討	21
4.3	區塊擴張的探討	31
4.4	比較與討論	37
第五章	結論	44
參考文獻	45
附錄-英文論文   48
 
圖目錄
圖1.1 複製移動竄改影像   3
圖2.1 前人方法   9
圖2.2  DOG與極值   11
圖2.3 高斯金字塔   12
圖2.4 特徵點方向性   16
圖2.5 特徵點描述示意圖   17
圖3.1 技術探討   19
圖4.1 不變動量值之比較   21
圖4.2 模糊處理下之個數   23
圖4.3 模糊處理下之時間   23
圖4.4 模糊處理下之正確率   24
圖4.5 模糊處理下之偵測率   24
圖4.6 模糊處理下之結果圖   27
圖4.7 關鍵點配對之組數   28
圖4.8 關鍵點配對之時間   28
圖4.9 關鍵點配對之正確率   29
圖4.10 關鍵點配對之偵測率   29
圖4.11 關鍵點配對之結果圖   31
圖4.12 區塊擴張之時間   33
圖4.13 區塊擴張之正確率   34
圖4.14 區塊擴張之偵測率   34
圖4.15 區塊擴張之結果圖   36
圖4.16 誤差區塊示意圖   37
圖4.17 硬幣圖之實驗分布圖   39
圖4.18 鳥群圖之實驗分布圖   39
圖4.19 羊群圖之實驗分布圖   39
圖4.20 相似度比較   42
圖4.21 較大複製區塊   42
圖4.22 最佳參數之結果圖   43
 
表目錄
表4.1 最佳參數   40
表4.2 最佳參數之結果數據   40
表4.3 最佳模糊選取   41
表4.4 相似度比較   42

[1] B. Ustubioglu, V. Nabiyev, G. Ulutas and M. Ulutas, “Image forgery detection using colour moments,” International Conference on Telecommunications and Signal Processing (TSP), 2015.

[2] C. S. Lin, C. C. Chen and Y. C. Chang. “An efficiency enhancedcluster expanding block algorithm for copy-move forgery etection,”International Conference on Intelligent Networking and Collaborative Systems (INCOS), 2015. 

[3] C. C. Chen, L. Y. Chen and Y. J. Lin, “Block Sampled Matching with Region Growing for Detecting Copy-Move Forgery Duplicated Regions,” Journal of Information Hiding and Multimedia Signal Processing, pp. 86-96, 2017.

[4] D.G.Lowe.“Distinctive image features from scale-invariant keypoints”International Journal of Computer Vision, vol. 60, no.2, pp.91-110,Nov. 2004.

[5] G. Emre, U. Güzin and U. Mustafa, “Rotation invariant copy move forgery detection method,” International Conference on Electrical and Electronics Engineering (ELECO), 2015.

[6]G. Lynch, F. Y. Shih and H. M. Liao, “An efficient expanding block algorithm for image copy-move forgery detection,” Information Sciences, 239, pp. 253-265, 2013.

[7]H. K. Tu, L. T. Thuong, H. V. U.Synh and H. V. Khoa, “The efficiency of applying DWT and feature extraction into copy-move images detection,” International Conference on Advanced Technologies for Communications (ATC), 2015.

[8]L. Li, S. Li, H. Zhu and X. Wu, “Detecting copy-move forgery under affine transforms for image forensics,” Computers and Electrical Engineering, 40(6), pp. 1951-1962, 2014.

[9] M. K. Hu, “Visual pattern recognition by moment invariants,” IRETransactions on Information Theory 8(2), pp. 179-187, 1962.

[10]Mikolajczyk, K., and Schmid, C. 2002. “An affine invariant interest point detector.” In European Conference on Computer Vision (ECCV), Copenhagen, Denmark, pp. 128-142.

[11] O. M. Al-Qershi and B. E. Khoo, “Passive detection of copy-move forgery in digital images: State-of-the-art,” Forensic Science International, 231(1–3), pp. 284-295, 2013.

[12] R. Davarzani, K. Yaghmaie, S. Mozaffari and M. Tapak, “Copy-move forgery detection using multiresolution local binary patterns,”Forensic Science International, 231(1–3), pp. 61-72, 2013.

[13] S. X. Liao, M. Pawlak.“On the accuracy of Zernike moments for image analysis”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol.20, pp.1358-1364, 1998

[14] S. J. Ryu, M. Kirchner, M. J. Lee and H. K. Lee, “Rotation invariant localization7 of duplicated image regions based on Zernike moments,” IEEE Transactions on Information Forensics and Security 8(8), pp.1355-1370, 2013.

[15] S. Y. Liao and T. Q. Huang, “Video copy-move forgery detection and localization based on Tamura texture features,” International Congress on Image and Signal Processing (CISP), 2013.

[16] V. Subramanyam and S. Emmanuel, “Pixel estimation based video forgery detection,” IEEE International Conference on Acoustics,Speech and Signal Processing, 2013.

[17] X. Bi, C. Pun and X. Yuan, “Multi-level dense descriptor and hierarchical feature matching for Copy–Move forgery detection,” Information Sciences, vol. 345, pp. 226-242. 2016.

[18]Y. Li, “Image copy-move forgery detection based on polar cosine transform and approximate nearest neighbor searching,” Forensic Science International, 224(1–3), pp. 59-67, 2013.