Real-Time Color Image Watermarking Based on D-SVD Scheme
This investigation presents a robust digital watermarking scheme for copyright protection, called D-SVD. The proposed scheme integrates discrete cosine transform (DCT) with singular value decomposition (SVD). In contrast to traditional methods based on DC
- PDF / 1,130,694 Bytes
- 9 Pages / 430 x 660 pts Page_size
- 9 Downloads / 260 Views
Abstract. This investigation presents a robust digital watermarking scheme for copyright protection, called D-SVD. The proposed scheme integrates discrete cosine transform (DCT) with singular value decomposition (SVD). In contrast to traditional methods based on DCT watermarking schemes, in which watermark messages are embedded directly on DCT frequency coefficients, the proposed approach emphasizes that watermark message bit is embedded on the blocks of the DCT coefficient’s singular value within an original color image. Experimental results demonstrate that the quality of watermarked image is robust under compression, noises, filtering and various attacks. In addition, it is observed that the proposed D-SVD algorithm can obtain larger NC and PSNR values than some existing well-known methods, and can successfully resist attacks such as cropping, blurring, reshaping, adding noise and JPEG compression. Keywords: digital watermarking, copyright protection, image processing.
1
Introduction
Digital watermarking algorithms are generally developed to verify authorship or to protect copyright [1], [4]-[18]. Numerous watermarking schemes that embed the watermark into transformed frequency coefficients have been presented recently, due to the robustness consideration such as DCT [1], [3], [13], [16] and wavelet [2], [18]. This methodology is unlike conventional approaches based on the transformed coefficient. This study proposes a new method that embeds the watermark into the singular value of the DCT coefficient blocks within an original color image, and does not require the original image when performing extraction of watermark procedure. The proposed scheme first calculates the frequency coefficient of luminance within the original color image using DCT, and decomposes the coefficient to computes its singular value after. The watermarks are then embedded on the singular value. Experimental results indicate that the presented technique produces a high-quality watermarked image and a robust embedding watermark. D. Mery and L. Rueda (Eds.): PSIVT 2007, LNCS 4872, pp. 289–297, 2007. c Springer-Verlag Berlin Heidelberg 2007
290
C.-F. Tsai and W.-Y. Yang
Fig. 1. The proposed watermark embedding process
2
Embedding Process
According to illustrated in Fig. 1, the proposed process embeds an n × n watermark image W into an N × N size original image H, with secret keys (key) for security. The steps of watermark embedding process are provided as follows: Step 1: Transform the RGB color space of the original image H to the YUV color space. The Y (luminance) component has selected from YUV to apply the DCT frequency transformation, since the luminance has the same range as the gray image between 0 and 255. The original image luminance component Y is split into non-overlapping 8 × 8 or 4 × 4 pixel blocks Bk , where k=1, 2, 3,. . . , M , and M represents the numbers of blocks. Each block Bk is then individually transformed to a frequency coefficient using the two-dimension DCT. Step 2: Compute the singular value of each frequency coefficient
Data Loading...
