Author Affiliations
1College of Information Science and Engineering, Yanshan University, Qinhuangdao, Hebei 0 66004 China2School of Electronics Engineering and Computer Science, Peking University, Beijing 100871, China3College of Software, Liaoning Technical University, Huludao, Liaoning 125105, China4Beijing Engineering Technology Research Center of Virtual Simulation and Visualization, Peking University, Beijing 100871, China5Key Laboratory for Computer Virtual Technology and System Integration of Hebei Province, Qinhuangdao, Hebei 0 66004show less
Fig. 1. Preprocessing process of copyright watermark
Fig. 2. Process of constructing zero watermark and extracting copyright watermark. (a) Generate zero watermark; (b) extract copyright watermark
Fig. 3. Carrier images and copyright watermark. (a) Baboon image; (b) barbara image; (c) boat image; (d) couple image; (e) elaine image; (f) goldhill image; (g) lena image; (h) paper image; (i) sailboat image; (j) copyright logo; (k) watermark mapped; (l) QR code
Fig. 4. Relationship between the change rate of maximum singular value and modulation factors. (a) Gaussian noise (0.3); (b) median filtering (template size 11×11); (c) JPEG compression (compression factor 10); (d) rotation attack (5°); (e) column and column offset (5 rows down); (f) crop attack (upper left 1/4)
Fig. 5. Combined attack. (a) Gaussian noise+rotation attack; (b) shear attack+Gaussian noise; (c) salt and pepper noise+JPEG compression; (d) median filter+rotation attack; (e) rotation attack+row and column offset; (f) shear attack+JPEG
Fig. 6. Zero watermark security test. (a) Baboon image; (b) barbara image; (c) boat image; (d) couple image; (e) elaine image; (f) goldhill image; (g) lena image; (h) paper image; (i) sailboat image
Fig. 7. Comparison results of different algorithms. (a) Gaussian noise; (b) median filter; (c) JPEG compression; (d) rotation attack; (e) row and colum offset; (f) crop position
Item | Baboon | Barbara | Boat | Couple | Elaine | Goldhill | Lena | Paper | Sailboat |
---|
Baboon | 1.0000 | 0.4981 | 0.7067 | 0.6744 | 0.7280 | 0.5317 | 0.6151 | 0.6963 | 0.6087 | Barbara | 0.4981 | 1.0000 | 0.5504 | 0.6036 | 0.5129 | 0.6498 | 0.5856 | 0.5682 | 0.6564 | Boat | 0.7067 | 0.5504 | 1.0000 | 0.7372 | 0.7297 | 0.7150 | 0.7157 | 0.6268 | 0.6793 | Couple | 0.6744 | 0.6036 | 0.7372 | 1.0000 | 0.6855 | 0.7135 | 0.7638 | 0.6725 | 0.7280 | Elaine | 0.7280 | 0.5129 | 0.7297 | 0.6855 | 1.0000 | 0.6703 | 0.7372 | 0.6538 | 0.6391 | Goldhill | 0.5317 | 0.6498 | 0.7150 | 0.7135 | 0.6703 | 1.0000 | 0.7274 | 0.5915 | 0.7355 | Lena | 0.6151 | 0.5856 | 0.7157 | 0.7638 | 0.7372 | 0.7274 | 1.0000 | 0.5604 | 0.7640 | Paper | 0.6963 | 0.5682 | 0.6268 | 0.6725 | 0.6538 | 0.5915 | 0.5604 | 1.0000 | 0.5219 | Sailboat | 0.6087 | 0.6564 | 0.6793 | 0.7280 | 0.6391 | 0.7355 | 0.7640 | 0.5219 | 1.0000 |
|
Table 1. Experiment results of false alarm rate
Attack type | Parameter | Baboon | Barbara | Boat | Couple | Elaine | Goldhill | Lena | Paper | Sailboat |
---|
Salt & pepper noise | 0.01 | 0.9948 | 0.9983 | 1.0000 | 1.0000 | 1.0000 | 0.9974 | 0.9983 | 0.9974 | 0.9965 | 0.05 | 0.9878 | 0.9913 | 0.9931 | 0.9861 | 0.9948 | 0.9895 | 0.9983 | 0.9922 | 0.9957 | Gaussian noise | 0.01 | 0.9887 | 0.9931 | 0.9930 | 0.9800 | 0.9922 | 0.9861 | 0.9931 | 0.9904 | 0.9930 | 0.05 | 0.9809 | 0.9843 | 0.9825 | 0.9660 | 0.9835 | 0.9801 | 0.9844 | 0.9869 | 0.9737 | Speckle noise | 0.01 | 0.9931 | 0.9983 | 0.9965 | 0.9948 | 1.0000 | 0.9948 | 0.9983 | 0.9974 | 1.0000 | 0.05 | 0.9861 | 0.9983 | 0.9930 | 0.9896 | 0.9965 | 0.9930 | 0.9930 | 0.9931 | 0.9896 | Median filtering | 3×3 | 0.9948 | 0.9965 | 1.0000 | 0.9948 | 1.0000 | 0.9983 | 0.9965 | 1.0000 | 0.9895 | 9×9 | 0.9801 | 0.9913 | 0.9853 | 0.9878 | 0.9965 | 0.9965 | 0.9913 | 0.9931 | 0.9825 | Wiener filtering | 3×3 | 0.9965 | 0.9983 | 1.0000 | 0.9983 | 1.0000 | 0.9965 | 1.0000 | 1.0000 | 0.9930 | 9×9 | 0.9931 | 0.9913 | 0.9913 | 0.9965 | 1.0000 | 0.9965 | 0.9948 | 0.9965 | 0.9895 | Low pass filtering | 3×3 | 0.9913 | 1.0000 | 0.9965 | 0.9983 | 0.9983 | 0.9965 | 0.9983 | 0.9965 | 0.9930 | 9×9 | 0.9913 | 0.9965 | 0.9948 | 0.9983 | 0.9983 | 0.9965 | 0.9983 | 0.9965 | 0.9860 | JPEG compression | 10 | 1.0000 | 0.9965 | 0.9913 | 1.0000 | 0.9965 | 0.9939 | 0.9965 | 0.9983 | 0.9965 | 50 | 1.0000 | 1.0000 | 0.9965 | 0.9983 | 1.0000 | 0.9983 | 0.9983 | 0.9974 | 1.0000 |
|
Table 2. Experimental results of non-geometry attack
Attack type | Item | Baboon | Barbara | Boat | Couple | Elaine | Goldhill | Lena | Paper | Sailboat |
---|
Rotation /(°) | 1 | 0.9776 | 0.9819 | 0.9755 | 0.9704 | 0.9754 | 0.9755 | 0.9836 | 0.9801 | 0.9577 | 2 | 0.9538 | 0.9512 | 0.9549 | 0.9567 | 0.9533 | 0.9613 | 0.9837 | 0.9625 | 0.9376 | 5 | 0.9036 | 0.9148 | 0.9109 | 0.8998 | 0.8755 | 0.8977 | 0.9377 | 0.8811 | 0.9038 | Cutting | Upper left 1/4 | 0.9392 | 0.9705 | 0.9396 | 0.9314 | 0.9447 | 0.9471 | 0.9555 | 0.9623 | 0.9755 | Center 1/4 | 0.9368 | 0.9623 | 0.9447 | 0.9461 | 0.9633 | 0.9639 | 0.9515 | 0.9441 | 0.9650 | Bottom right 1/4 | 0.9498 | 0.9621 | 0.9416 | 0.9594 | 0.9314 | 0.9737 | 0.9602 | 0.9657 | 0.9477 | Move | 1 line | 0.9948 | 0.9878 | 0.9939 | 0.9913 | 1.0000 | 0.9930 | 0.9983 | 0.9861 | 0.9965 | 2 line | 0.9931 | 0.9843 | 0.9869 | 0.9861 | 0.9983 | 0.9792 | 0.9965 | 0.9843 | 0.9869 | 5 column | 0.9514 | 0.9586 | 0.9755 | 0.9574 | 0.9684 | 0.9531 | 0.9406 | 0.9492 | 0.9489 |
|
Table 3. Experimental results of geometry attack
Attack type | Parameter | SVD | ASVD |
---|
NC | BER /% | NC | BER /% |
---|
Gaussian noise | 0.2 | 0.3777 | 52.34 | 0.9913 | 0.98 | Median filtering | 5×5 | 0.6923 | 30.57 | 1.0000 | 0.00 | JPE compression | 20 | 0.5952 | 39.16 | 0.9983 | 0.20 | Rotation /(°) | 2 | 0.3058 | 53.91 | 0.9575 | 4.79 | Cutting | Upper left1/4 | 0.9121 | 9.47 | 0.9648 | 3.91 | Move | 5 line | 0.3097 | 54.39 | 0.9895 | 1.17 |
|
Table 4. Performance comparison between SVD and ASVD