Multipurpose Watermarking Algorithm for Medical Images

Table 2

Similarity of watermark extracted under various attacks.

Attack types	Similarity of robust watermark (NC)
	Skull					Mediastinum					Liver
	This paper	[14]	[27]	[26]	[28]	This paper	[14]	[27]	[26]	[28]	This paper	[14]	[27]	[26]	[28]
No attack	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1
Gaussian filter (3  3  = 0.3)	0.956	—	0.979	—	0.971	0.954	—	0.977	—	0.967	0.944	—	0.972	—	0.959
Median filter [3, 3]	0.942	0.956	0.987	0.861	0.983	0.938	0.951	0.985	0.856	0.982	0.926	0.944	0.982	0.847	0.98
White noise (0.01)	0.961	—	—	0.970	0.958	0.959	—	—	0.969	0.956	0.951	—	—	0.963	0.949
Salt and pepper noise (0.02)	0.957	0.940	0.949	0.946	0.942	0.961	0.942	0.947	0.944	0.940	0.952	0.935	0.942	0.938	0.937
Shear (1/16)	0.921	0.852	—	0.935	0.941	0.920	0.863	—	0.931	0.933	0.906	0.842	—	0.920	0.922
JPEG compression Q = 20	0.914	—	—	—	0.743	0.934	—	—	—	0.755	0.921	—	—	—	0.728
JPEG compression Q = 50	0.955	—	0.872	0.822	0.857	0.952	—	0.883	0.820	0.870	0.943	—	0.865	0.809	0.854
Shrink 30%	0.982	0.977	—	0.861	1	0.985	0.978	—	0.858	1	0.976	0.971	—	0.857	1
Enlarge 30%	0.985	0.971	—	0.853	1	0.988	0.973	—	0.855	1	0.981	0.968	—	0.849	1
Rotate 30°C	0.983	—	—	0.892	0.975	0.978	—	—	0.888	0.974	0.972	—	—	0.883	0.972
Rotate 45°C	0.976	0.967	—	—	0.964	0.972	0.965	—	—	0.963	0.970	0.961	—	—	0.959