Watermark Detection and Extraction Using Independent Component Analysis Method
- PDF / 6,636,149 Bytes
- 13 Pages / 600 x 792 pts Page_size
- 39 Downloads / 203 Views
atermark Detection and Extraction Using Independent Component Analysis Method Dan Yu School of Electrical and Electronic Engineering, Nanyang Technological University, Nanyang Avenue, Singapore 639798 Email: [email protected]
Farook Sattar School of Electrical and Electronic Engineering, Nanyang Technological University, Nanyang Avenue, Singapore 639798 Email: [email protected]
Kai-Kuang Ma School of Electrical and Electronic Engineering, Nanyang Technological University, Nanyang Avenue, Singapore 639798 Email: [email protected]. Received 30 July 2001 and in revised form 12 October 2001 This paper proposes a new image watermarking technique, which adopts Independent Component Analysis (ICA) for watermark detection and extraction process (i.e., dewatermarking). Watermark embedding is performed in the spatial domain of the original image. Watermark can be successfully detected during the Principle Component Analysis (PCA) whitening stage. A nonlinear robust batch ICA algorithm, which is able to efficiently extract various temporally correlated sources from their observed linear mixtures, is used for blind watermark extraction. The evaluations illustrate the validity and good performance of the proposed watermark detection and extraction scheme based on ICA. The accuracy of watermark extraction depends on the statistical independence between the original, key and watermark images and the temporal correlation of these sources. Experimental results demonstrate that the proposed system is robust to several important image processing attacks, including some geometrical transformations—scaling, cropping and rotation, quantization, additive noise, low pass filtering, multiple marks, and collusion. Keywords and phrases: watermarking, dewatermarking, independent component analysis (ICA).
1. INTRODUCTION Digital watermarking technology has evolved very quickly these years. The basic principles of most watermarking methods are applying small, pseudorandom changes to the selected coefficients in the spatial or transform domain. Most of the watermark detection schemes use some kinds of correlating detector to verify the presence of the embedded watermark [1, 2]. The watermark can be extracted with information of the key, and with/without the original (i.e., unwatermarked) image. Independent Component Analysis (ICA) is probably the most powerful and widely-used method for performing Blind Source Separation (BSS). It is a very general-purpose statistical technique to recover the independent sources given only sensor observations that are linear mixtures of independent source signals [3, 4, 5]. The simplest BSS model assumes the existence of n independent components s1 , s2 , . . . , sn , and the same number of linear and instantaneous mixtures of these
sources, x1 , x2 , . . . , xn , that is, xj = aj1 s1 + aj2 s2 + · · · + ajn sn ;
1 ≤ j ≤ n.
(1)
In vector-matrix notation, the above mixing model can be represented as x = As,
(2)
where A is the square n × n mixing matrix. The unmixing process [3, 4, 5] can be formulated as
Data Loading...
