Tuning Parameters of the Koch and Zhao Stego Algorithm

Abstract

This paper assumes that the Laplacian model is valid for the middle frequency DCT coefficients of a JPEG image and considers embedding with a two coefficient version of the Koch and Zhao stego algorithm. Under some mederate conditions, one can get a dependency of the cover image PSNR vs payload and embedding threshold. Employing some algebra allows us to substitute this bulk expression for an approximate but simple one. Next, solving the resulting equation for the embedding threshold, given PSNR and payload, yields us an approximate value of the threshold. On the other hand, by solving numerically the original, non-simplified equation, we get the math-model-accurate value of the threshold. Thus, it becomes possible to compare two threshold values (accurate and approximate). This comparison has been done for different texture-like images and results are listed below.