Laser & Optoelectronics Progress, Volume. 57, Issue 10, 101102(2020)
Image Tampering Detection Method Based on Approximate Nearest Neighbor Search
Figures & Tables(12)
Fig. 1. Feature matching steps. (a) Initialization; (b) propagation; (c) random search
Fig. 2. Detection results of the algorithm on rotation transform operation. (a) Original image; (b) tampering image (rotation 90°); (c) matching result; (d) detection result
Fig. 3. Detection results of the algorithm on scaling transformation operation. (a) Original image; (b) tampering image (reduced by 80%); (c) matching result; (d) detection result
Fig. 4. Detection results of the algorithm on mirror transformation operation. (a) Original image; (b) mirror image (horizontal); (c) matching result; (d) detection result
Fig. 5. Detection results of the algorithm for multi-region I transform operation. (a) Original image; (b) tampering image; (c) matching result; (d) detection result
Fig. 6. Detection results of the algorithm for multi-region II transform operation. (a) Original image; (b) tampering image; (c) matching result; (d) detection result
Fig. 7. Detection results of the algorithm for multi-region mirroring I transform operation. (a) Original image; (b) tampering image; (c) matching result; (d) detection result
Fig. 8. Detection results of the algorithm for multi-region mirroring II transform operation. (a) Original image; (b) tampering image; (c) matching result; (d) detection result
Table 1. Experimental parameters
View tableTable 1. Experimental parameters
Parameter Value Explanation Td 50 Minimum distance of tampered area τ 360 Minimum threshold of minimum square linear model Tm 1000 Minimum radius of tampered area ρm 5 Radius of median filter ρe 8 Regional radius of minimum square linear model
Table 2. F-measure of algorithm detection under different tampering operations%
View tableTable 2. F-measure of algorithm detection under different tampering operations%
Algorithm Rotate 90° Zoom 80% Multi- regionⅠ Multi- region Ⅱ Ref. [4] 81.34 50.14 54.15 62.12 Ref. [2] 79.17 84.23 79.31 84.13 Ref. [6] 85.54 83.54 91.23 92.15 This algorithm 94.76 90.45 93.75 93.12
Table 3. Comparison of average running time of different algorithmss
View tableTable 3. Comparison of average running time of different algorithmss
Algorithm Rotate Zoom Multi- regionⅠ Multi- regionⅡ Ref. [4] 39.87 77.49 507.45 511.94 Ref. [2] 53.91 56.84 110.67 146.78 Ref. [6] 34.02 43.62 303.87 297.59 This algorithm 27.28 33.19 72.94 81.85
Table 4. Comparison of tampering detection performance of different algorithms under mirror operation
View tableTable 4. Comparison of tampering detection performance of different algorithms under mirror operation
Algorithm F-measure Time /s Mirror image /% Multi mirror image /% Ref. [11] 90.62 53.46 140.43 Ref. [12] 84.43 72.90 165.72 Ref. [13] 91.12 75.12 179.32 This algorithm 94.49 93.52 80.97
Tools
Get Citation
Copy Citation Text
Jing Wang, Yuchen Zhang, Zhanqiang Huo, Liqin Jia. Image Tampering Detection Method Based on Approximate Nearest Neighbor Search[J]. Laser & Optoelectronics Progress, 2020, 57(10): 101102
Paper Information
Category: Imaging Systems
Received: Aug. 13, 2019
Accepted: Oct. 11, 2019
Published Online: May. 8, 2020
The Author Email: Zhanqiang Huo (hzq@hpu.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20