A New Image Watermarking Algorithm Using the Contourlet Transform and the Harris Detector
In this paper, we propose a new feature-based image watermarking scheme based on multiscale theory and the Contourlet transform (CT). We use the multiscale Harris detector to extract stable feature points from the host image. Next, according to feature sc
- PDF / 210,367 Bytes
- 9 Pages / 439.37 x 666.14 pts Page_size
- 87 Downloads / 211 Views
College of Information Science and Engineering, Northeastern University, Shenyang 110819, China [email protected] 2 Department of Computer Science, Tonghua Normal University, Tonghua 134000, China 3 College of Computer Science and Technology, Taiyuan University of Technology, Taiyuan 030024, China
Abstract. In this paper, we propose a new feature-based image watermarking scheme based on multiscale theory and the Contourlet transform (CT). We use the multiscale Harris detector to extract stable feature points from the host image. Next, according to feature scale theory, we determine the local feature regions (LFR) and scale the regions to a standard size. We then embed the digital watermark into the Contourlet low frequency area calculated using the pseudo-Zernike moment. The results of our experiments demonstrate that the algorithm results in an invisible watermark and is robust against conventional signal processing (median filtering, sharpening, noise adding, and JPEG compression), geometric attacks (rotation, translation, scaling, row or column removal, shearing, local geometric distortion) and combined attacks. Keywords: Image watermarking · Geometric attacks · Contourlet transform · Pseudo-Zernike moments · Low sub-band1
1
Introduction
Digital watermarking is widely used in Internet multimedia intellectual property protection [1-2]. Great progress has been achieved in the image transform domain by applying the discrete cosine transform (DCT)[3-4] and discrete wavelet transform (DWT)[5-6] in digital watermarking algorithms. However, these theories do not adequately represent the anisotropy of signals. The Ridgelet [7], Curvelet [8], and Concourlet [9] transforms better address the anisotropy of signals and can effectively resist Random bending attack (RBA), Geometric attacks (i.e., Rotation, Scale and Translation, RST), and Shearing. Many methods have been proposed to evade geometric attacks. These methods can be roughly divided into three categories: (1) invariant transforms [12-13], which are robust against global geometrical distortion but not against shearing; (2) feature-based synchronization [14-15], whose watermarking capacity is very limited; and (3) template insertion [16], which cannot withstand any malicious attacks. © Springer-Verlag Berlin Heidelberg 2015 H. Zha et al. (Eds.): CCCV 2015, Part II, CCIS 547, pp. 439–447, 2015. DOI: 10.1007/978-3-662-48570-5_42
440
D. Zhu and L. Lv
To address the above prroblems, we present a new digital watermarking scheme, which includes two main feeatures. First, an analysis of Contourlet transform characcteristics indicates that the pro oposed scheme can adaptively embed a watermark into llow frequency sub-bands with many m textures. Second, we review feature-based synchroonization methods and proposse a novel watermark synchronization strategy. In expperiments, we compare the perrformance of the proposed algorithm with other algorithhms by applying various attack ks. The results demonstrate the superior robustness of the proposed watermarking algorithm.
2
Data Loading...
