Digital Image Watermarking in DCT Domain Using Finite Newton Support Vector Regression

Finite Newton Support Vector Regression (FNSVR) is a reliable and robust method for regression analysis which unlike other support vector regression methods, converges the computation in a few iterations. In this paper, we have used FNSVR method for embedding and extracting a binary image as a watermark in 8-bit gray scale cover images. Digital image watermarking being a time consuming task, requires fast embedding and extraction procedures and overall it should have a low time complexity. FNSVR being an algorithm which consumes less number of iterations, embeds a given watermark in a short time span. Computed PSNR values indicate good quality of the signed images. SIM(X,X^*) parameter is computed and its values indicate that the extraction process is quite successful.