A Novel Video Watermarking Technique Based On Multiresolution Singular Value Decomposition NOUIOUA Imen Laboratoire Croissance et Caractérisation de Nouveaux Semi-conducteurs (LCCNS) – Département d'Electronique - Université FERHAT Abbas de Sétif imen_bou84@hotmail.fr AMARDJIA Nourredine Laboratoire Instrumentation Scientifique (LIS) – Département d'Electronique Université FERHAT Abbas de Sétif amardjia_nour@yahoo.fr Abstract— In this paper, a new robust video watermarking technique is proposed which combines Singular Value Decomposition (SVD) and its multiresolution form (MR-SVD). The experimental results obtained through the proposed scheme demonstrate that this technique provides imperceptible watermarks and is robust against several attacks like noise addition, rotation and MPEG compression. Keywords- Video watermarking; Singular Value Decomposition (SVD); Multiresolution-SVD (MR-SVD). I. INTRODUCTION Due to the extraordinary revolution of internet, the digital multimedia (movie, music, and image) can be easily obtained with low cost and high quality. Therefore, owners and creators of digital products are concerned about illegal copying of their products. The Digital Watermarking has been proposed as a solution to the problem of copyright protection [1]. The basic idea of the “watermarking” is to embed a robust and subliminal (invisible or inaudible) data into multimedia elements [2]. Watermarking techniques can be classified into two classes: spatial domain and frequency domain. The spatial domain methods embed the watermark by modifying the pixel values of the host directly, but in the frequency domain techniques, the watermark is embedded by modifying the transform coefficients of the host .The most commonly used transforms are the Discrete Fourier Transform (DFT), the Discrete Cosine Transform (DCT), the Discrete Wavelet Transform (DWT) and the singular value decomposition (SVD) [3, 4]. Kakarala and Ogunbona [5] have proposed a multiresolution form of the SVD (MR-SVD) and showed how it could be used for signal analysis. A new video watermarking scheme is presented here, where the SVD is combined with the MR-SVD. This paper is organized as follow. To understand our scheme, we introduce in the next section the concept of SVD and MR-SVD transformation techniques. In section three, we propose our video watermarking method where we describe the watermark embedding and extraction processes. In section four, we present the experimental results and in section five, we compare our results with existing work done by previous researchers. The conclusions of our study are stated in section six. II. THEORETICAL FUNDAMENTALS A. Singular Value Decomposition Singular Value Decomposition is said to be a significant topic in linear algebra by renowned mathematicians. Also SVD has shown its usefulness in variety of applications including image processing and watermarking. If A is a matrix, representing for example an image of size m*n, then the SVD of A is given by: T A U*S*V = (1) Where U and V are orthogonal matrices and S is a diagonal matrix { } ( ) 1 2 n S diag s ,s s = 4 s 1 , s 2 … s n are the singular values (SV) of A, which observe the condition s 1 œ s 2 œ….œ s n ; These SV represent the energy of the image and have very good stability [6]. B. Multiresolution singular value decomposition The MR-SVD represents a signal as a series of an approximation and details like the DWT. This section explains how the MR-SVD, introduced in [5], works. 1) 1D Multiresolution Singular Value Decomposition Let X=[x(1) … x(N)] represent a finite extent 1-D signal. Assume that N is divisible by 2L for some L œ1. Let the data matrix at the first level, denoted X1, be constructed so that its top row contains the odd-numbered samples and the bottom row contains the even-numbered samples: ( ) ( ) ( ) 1 x1 x3 xN 1 X x(2) x(4) x(N) −   =     4 4 (2) Let U 1 be the eigenvector matrix bringing the scatter matrix T 1 =X 1 T 1 X into diagonal form: T 2 1 1 1 1 U TU S = (3) Where S 1 2 =diag{s 1 (1) 2 ,s 1 (2) 2 } contains the squares of the two singular values, with s 1 (1) œ s 1 (2). Now let: Proceedings of The first International Conference on Nanoelectronics, Communications and Renewable Energy 2013 323 ICNCRE ’13 ISBN : 978-81-925233-8-5 www.edlib.asdf.res.in Downloaded from www.edlib.asdf.res.in