Video Dual Watermarking Algorithm against Geometric Attack Based on ASIFT
and Contourlet Transform
Shuqin Chen
School of Computer Science and Technology
Guizhou University
Guiyang, China
e-mail: csqcwx@foxmail.com
Zhi Li
School of Computer Science and Technology
Guizhou University
Guiyang, China
e-mail: lizhigzu@163.com
Xinyu Cheng
School of Computer Science and Technology
Guizhou University
Guiyang China
e-mail: xycheng@gzu.edu.cn
Qi Gao
School of Computer Science and Technology
Guizhou University
Guiyang, China
e-mail: gq813@foxmail.com
Abstract—This study proposed a video dual watermarking
algorithm based on affine-scale invariant feature transform
(ASIFT) and contourlet transform. First, the human visual
masking model of a 3D motion in video sequence is studied in
depth. The human eye visual masking threshold is obtained as
the maximum embedding intensity of watermark using various
motion characteristics. Second, the high- and low-frequency
sub-band coefficients of the contourlet field are obtained by
contourlet transform. Chaotic watermarking sequence is
embedded into the high-frequency sub-band coefficient with
the highest energy to increase imperceptibility. Third, when
the low-frequency sub-band coefficients has the stability of its
coefficient histogram against geometric attacks such as
rotation and scaling, the watermark signal is embedded in a
low-frequency sub-band histogram of adjacent coefficients to
increase the watermark of an anti-geometric attack. Finally,
ASIFT is used as a trigger to determine whether the video
frame is subjected to geometric attacks or not. For geometric
distortions, ASIFT is used to regulate the geometrically
attacked video frame. The low-frequency sub-band coefficients
of the regulated video frame are used for the watermarking
extraction algorithm. The high-frequency watermarking
extraction algorithm is used directly for the non-geometric
distortions. Experimental results show that the proposed
algorithm could guarantee watermark invisibility and
favorably extract the watermark for common geometric and
conventional signal attacks. The proposed algorithm is a strong
video-dual watermarking algorithm.
Keywords-anti-geometric attack; dual watermarking; ASIFT;
contourlet transform
I. INTRODUCTION
With the development of Internet and information
technology, digital watermarking technology, which could
protect the copyright of digital multimedia work and ensure
information security, has become a hot spot in the research
field of information. Digital watermarking requirements
include invisibility, robustness, detectability and versatility,
where robustness and invisibility are contradictory
requirements.
In recent years, research on digital video watermarking
has made great progress. A series of digital video
watermarking algorithms has been proposed. Spatial domain
and transform algorithms are present in watermark
embedding algorithms. Although the spatial domain
algorithm is fast and easy to implement, its robustness is not
good. The transform domain algorithm usually has good
robustness. Thus, wavelet transform has local characteristics
in both time and frequency domains, where it is widely used.
In our previous study, we applied all the characteristics of
wavelet and integer wavelet transforms [1–2] and achieved
good results. However, the wavelet cannot be optimally
represented by the singularity of the higher dimension, such
as the curve or linear feature, in the 2D image. In 2002, Do
and Vetterli proposed a “true” 2D representation of the
image contourlet transform [3]. This new multi-scale
geometric transformation can entirely represent the
geometric characteristics of the image itself. Thus, the
watermarking algorithm based on contourlet transform is
widely used.
For example [4], the contourlet transform is applied in
the host image to obtain low- and high-frequency sub-bands.
The high-frequency sub-band uses multi-description coding
to embed the watermark into the odd descriptor. The low-
frequency sub-band is used to embed the watermark bit
through the quantization index modulation table. However,
the watermark is seriously damaged by several signal attacks,
such as median filter, salt-and-pepper noise, and Gaussian
noise. In [5], an adaptive robust digital watermarking
algorithm for the contourlet domain virtual tree structure is
proposed to coordinate the robustness and transparency of
the watermarking algorithm by using the fruit fly algorithm
to optimize the parameters of support vector regression.
However, the results of rotation, cutting, and other common
geometric attacks are not ideal. In [6], the watermarking
signal is synchronized with the scale-invariant feature
transform (SIFT), and a circular watermarking pattern is
1757
2017 17th IEEE International Conference on Communication Technology
978-1-5090-3944-9/17/$31.00 ©2017 IEEE