Color Image Watermarking using JND Sampling Technique
Rajesh B. Raut, Shri Ramdeobaba, Kishor M. Bhurchandi · Computer Engineering and Intelligent Systems · 2011
This paper presents a color image watermarking scheme using Just Noticeable Difference (JND) Sampling Technique in spatial domain. The nonlinear JND Sampling technique is based on physiological capabilities and limitations of human vision. The quantization levels have been computed using the technique for each of the basic colors R, G and B respectively for sampling color images. A watermark is scaled to half JND image and is added to the JND sampled image at known spatial position. For transmission of the image over a channel, the watermarked image has been represented using Reduced Biquaternion (RB) numbers. The original image and the watermark are retrieved using the proposed algorithms. The detection and retrieval techniques presented in this paper have been quantitatively benchmarked with a few contemporary algorithms using MSE and PSNR. The proposed algorithms outperform most of them. Digital watermarking is a technique to embed secret information in videos or images. Color image watermarking is a comparatively unexplored field relative to gray-level image watermarking. The watermarking algorithms can be either spatial domain or transform domain. Most of the watermarking schemes exploit limitations of the human vision i.e. the human eye sensitivity to the light spectrum, discrimination power for discriminating the colors etc. The sensitivity of Red color vision is the lowest, though it fascinates or attracts the most. The sensitivity of Green color vision is the highest. The Blue vision is moderately sensitive. Kutter, et. al. (1997) have used the blue channel of the RGB space for embedding the watermark as the human eye is moderately sensitive to blue color. The embedding process uses spread spectrum techniques in the frequency domain and then transforms it to spatial domain. Most of the existing algorithms mark the channels of color separately, without considering the correlation of the image or watermark color components. Reed & Hannigan (2002) proposed a system that takes advantage of the low sensitivity of the human visual system to high frequency changes along yellow-blue axis to place most of the watermark information in yellow component of the image in CMY space. The quality of watermarking should be such that it is not detected by the human eye or the sensitivity of vision to the light spectrum is very important while evaluating the quality of watermarking as presented as by Piva et. al. (1999). In this scheme, the embedding process is an additive (spread-spectrum) scheme in the DCT domain. The Lab color space decomposition presents interesting properties in the watermarking context and the same is used (Fleet and Heeger 1997). A watermarking scheme based on the QFT and the Quantization Index Modulation scheme is presented by Bas et. al. (2003). In this method, imperceptibility is emphasized but the robustness to extensive attacks is overlooked. Watermarking scheme for color images based on local quaternion Fourier spectral analysis (LQFSA) has been discussed in Xiaojun et. al. (2008). It uses Quaternion Fourier transform defined in a 4D vector space which provides a larger embedding scope for