![](https://csdnimg.cn/release/download_crawler_static/15708687/bg1.jpg)
1 INTRODUCTION
Pressed watermarked image as the outcome of digi-
tal watermark technology applying to anti-
counterfeiting or the printing media, has
drawn extensive attention due to its function of dis-
crimination and preventing piracy (Xie, 2013a, Xu,
2005, Chaitanya, 2013). When adopting water-
marked algorithm based on spatial domain and fre-
quency domain, it is inevitable to change the struc-
ture of host image due to embedding watermark and
cause the degradation of image quality (Zhang,
2012). In existing digital image watermark system,
most measure the difference between pre-embedding
and post-embedding images by peak signal to noise
ratio (PSNR) (Russo, 2008) parameter. But PSNR is
the global statistics based on the whole image so that
it can’t well identify the degree of degradation of the
local sensitive information, like biological eyes, bar
code, anti-counterfeiting pattern (curve underpaint-
ing), caused by watermark embed to the original
manuscript. We therefore have to find a new assess-
ment index adapted to local image quality degrada-
tion.
Aiming at continuous tone digital image quality
assessment, there exist assessment measurement like
MSE, PSNR, based on pixels' statistics theory PQS,
DVQ, based on HVS characteristic bionic theory and
UIQI (Wang, 2002), SSIM (Wang, 2004), WGM
based on feature parameter measurement. It remains
to validate whether these measures are suitable for
halftone pressing or/and printing image. Singular
Value Decomposition (SVD, (Modgil, 2012)) is ex-
tensively applied on digital image processing. Under
the consideration of SV, the singular value of the
two-dimensional image matrix decomposed by sin-
gular value decomposition, possessing the ability of
representing image local characteristics and SV’s
advantages of watermark image preprocessing, like
stability and rotation, translation and scale invari-
ance, etc. The paper proposes a sort of watermarked
printed image quality assessment measure based on
singular value. Compared with PSNR and UIQI,
which is more extensively applied in the field of dig-
ital image, referring to human visual perception,
comprehensively assess the printed matter.
2 THEORY AND METHOD
2.1 Image property base on SVD
As for any matrix A
Cm×n, its SVD (s
1
,s
2
,…
,
s
r
)
is unique. It describes the distribution characteristics
of matrix and shows nice image structural represen-
tation characteristics (Yuan, 2008). Suppose that A
is a digital image, the SVD of A can be described as:
11
0
00
rr
i i i i
ii
A USV U V A s u v
(1)
Quality evaluation of lithographic watermarked image based on singular
value
Y. Xie & H.H. Tan & K.L. Wang
Hunan University of Technology, Zhuzhou, Hunan, China
ABSTRACT: The digital watermarking technology application in the field of anti-counterfeiting printing, ap-
propriate image quality evaluation method is very important to optimize watermarking algorithm. Pressed im-
age is a special non-continuous halftone image. In order to better measure the effect of presswork cause by an-
ti-counterfeiting information embedded and to assess the transparence of watermarking algorithm, this article
presents a method based on singular value quality index (SVQI) to assessment the quality of pressed water-
marked image. The research result shows that comparing with PSNR and UIQI, SVQI is more sensitive to the
image details changes cause by watermark information embed and it’s quantitative analysis regularity more
obvious. Meanwhile experimental results show that SVQI is more suitable for evaluation halftone pressed wa-
termarked image.
KEYWORD: Image quality assessment; SVQI; SVD; PSNR; UIQI; Pressing watermarked image