Table 3 Tests on 10 copy-move manipulated images
From: Counter-forensics of SIFT-based copy-move detection by means of keypoint classification
| Â | Attack | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
Image(matches) | Classification-based | Smoothing | RMD | Collage | ||||||||
| Â | After (%) | Iter | PSNR | After (%) | Iter | PSNR | After (%) | Iter | PSNR | After (%) | Iter | PSNR |
I 1 (37) | 0 (100%) | 14 | 41.2 | 12 (67.6%) | 40 | 39.4 | 5 (86.5%) | 40 | 30.7 | 3 (91.9%) | 40 | 28.6 |
I 2 (66) | 0 (100%) | 20 | 39.9 | 29 (56.1%) | 40 | 47.7 | 21 (68.2%) | 40 | 25.6 | 6 (90.9%) | 40 | 33.7 |
I 3 (150) | 0 (100%) | 23 | 48.1 | 103 (31.3%) | 40 | 57.8 | 0 (100%) | 26 | 41.2 | 7 (95.3%) | 40 | 43.1 |
I 4 (41) | 0 (100%) | 15 | 42.8 | 11 (73.2%) | 40 | 42.1 | 9 (78.0%) | 40 | 27.4 | 2 (95.1%) | 40 | 32.2 |
I 5 (23) | 0 (100%) | 31 | 47.0 | 8 (65.2%) | 40 | 44.5 | 8 (65.2%) | 40 | 25.3 | 0 (100%) | 9 | 28.0 |
I 6 (56) | 0 (100%) | 17 | 45.8 | 27 (51.8%) | 40 | 41.1 | 7 (87.5%) | 40 | 26.3 | 3 (94.7%) | 40 | 33.3 |
I 7 (55) | 0 (100%) | 28 | 34.3 | 10 (81.8%) | 40 | 40.9 | 10 (81.8%) | 40 | 26.2 | 1 (98.2%) | 40 | 28.4 |
I 8 (32) | 0 (100%) | 15 | 45.0 | 8 (75.0%) | 40 | 38.7 | 1 (96.8%) | 40 | 30.1 | 2 93.8%) | 40 | 29.4 |
I 9 (52) | 0 (100%) | 19 | 45.8 | 21 (59.6%) | 40 | 48.6 | 10 (80.8%) | 40 | 25.3 | 2 (96.2%) | 40 | 33.5 |
I 10 (53) | 0 (100%) | 21 | 44.2 | 16 (69.8%) | 40 | 46.7 | 22 (58.5%) | 40 | 29.4 | 3 (94.3%) | 40 | 32.0 |
Average | 100% | 21 | 43.4 | 63.1% | 40 | 45.1 | 80.3% | 39 | 28.8 | 95.0% | 37 | 32.2 |