Watermarking of Grayscale Images in DCT Domain Using Least-Squares Support Vector Regression
Vikash Chaudhary, Anurag Kumar Mishra, Rajesh R. Mehta, Monika Verma, R. P. Singh, Navin Rajpal · International Journal of Machine Learning and Computing · 2012
In this paper, we have used Least-Squares Support Vector Regression (LS-SVR) method, which is a reliable and robust method for regression analysis, for grayscale image watermarking in DCT domain.This method offers several advantages unlike conventional SVRs which is perceived as a minimization problem with linear inequality constraints and has solution to quadratic programming (QP) problem.Due to this reason, the problem solution becomes computationally costly.On the other hand, the solution to the LS-SVR algorithm may be obtained by solving a system of linear equations instead of solving a QP problem and therefore it consumes less time.In this case, the LS-SVR algorithm embeds a given binary watermark in three different grayscale images in a short time span.PSNR values indicate good quality of the signed images.The watermarks are also extracted and the computed values of * ( , ) SIM X X correlation parameter indicate that the extraction process is quite successful.