In order to avoid the influence of some factors like tiredness or environment in the invisible watermarking evaluation via Human Visual System(HVS), this paper presents an efficient performance evaluation method on the imperceptibility, which is always consistant with human visual perception. In this method, after the original image and corresponding watermark-embedded images are decomposed by multilevel wavelet transform, their wavelet coefficients are divided into 4 categories in terms of the sensibility of HVS, and each of them is used to construct a reference sequence and some comparative sequences. The similarities of the two kinds of sequences are calculated by grey relational analysis, which is further adjusted with the consideration of the watermark-embedded-area ratio and the watermark-embedded coefficients. A global performance of imperceptibility of watermark-embedded images is obtained after the 4 grey relational degrees are combined with different weights. Experimental results show that compared to Peak Signal to Noise Ratio and some existing methods, the proposed method not only is efficient and universal, but also has a better ability to distinguish tiny details.
grey relational analysis,
Human Visual System(HVS),