Multi-bit Watermarking Scheme Based on Addition of Orthogonal Sequences - Computer Network Security - page 407

Multi-bit Watermarking Scheme
Based on Addition of Orthogonal Sequences
Xinpeng Zhang, Shuozhong Wang, and Kaiwen Zhang
Communication & Information Engineering, Shanghai University, Shanghai 200072, China
[e-mail],[e-mail],[e-mail]
Abstract. In this paper, a scheme of watermark embedding based on a set of or-
thogonal binary sequences is introduced. The described technique is intended to
be incorporated into various public watermarking frameworks developed for
different digital media including images and audio signals. Unlike some previ-
ous methods using PN sequences in which each sequence carries only one bit of
the watermark data, the proposed approach maps a number of bits to a single
sequence from an orthogonal set. Both analytical and experimental studies
show that, owing to the full exploitation of information carrying capability of
each binary sequence, the performance is significantly improved compared with
previous methods based on a one-bit-per-sequence technique.
1 Introduction
Watermark is a digital code embedded imperceptibly and robustly in the host data and
typically contains information about the owner, origin, status, and/or destination of
the data [1,2]. In terms of embedding capacity, watermarking schemes can be classi-
fied into single-bit and multiple-bit schemes. In single-bit schemes, the detection
results are simply binary, namely, “marked” or “not-marked”. It is often important for
the purpose of IPR protection, however, to embed more information into the host
signal such as the name and address of the owner. Therefore, a larger information
embedding capacity is desirable. Since there are conflicts between imperceptibility,
robustness and capacity, compromises must be sought.
Watermarking by adding pseudo-random (PN) sequences to the host data was em-
ployed in many techniques. The types of host include color images [3], audio signals
[4], vertex coordinates of polygonal models [5], etc. The method is applicable in spa-
tial domain [3,4], Fourier transform domain [5], DCT domain [6], wavelet domain [7-
9], and other domains [10-12]. To achieve better performances, various techniques
have been proposed. For example, watermark signals can be pretreated before inser-
tion to enhance robustness [4]. Adjusting watermark strength referenced to the host
data amplitude can provide better imperceptibility [7,8,11]. It is also helpful to adapt
the watermark strength to visual effects with respect to frequency locations and local
luminance [13]. Some other algorithms employ masking effects in spatial and/or fre-
quency domains when inserting watermarks [9,14,15].
Robustness of multi-bit watermarking schemes based on adding PN sequences
against different attacks has been studied [16]. However it is rather difficult for these
methods to provide a high embedding capacity since the sequences must be long
enough to ensure sufficient robustness. In view of this, other methods that are not