Image steganography based on difference of Gaussians edge detection

This paper introduces an edge-based image Steganography scheme in which the pixels of the cover images are categorized into two classes: edge and non-edge. In general, the edge pixels hide more secret bits compared to non-edge pixels due to the following two reasons: noisy nature and high tolerance level. The edge pixels are perceived as noisy due to the variation in intensities with respect to the neighboring pixels and hence it is difficult to model. Further the tolerance level of edge pixels is usually high compared to non-edge pixels on equivalent alteration. These two reasons motivate us to propose a new image Steganography method based on Difference of Gaussians (DoG) Edge detection. The proposed scheme has three major phases: pre-processing cum edge detection, embedding and extraction. In the leading two phases, we obtain the edge image from the cover image and then embed secret bits into the non-edge and edge pixels with X:Y ratio, where for all X, Y, 1≤X≤3and2≤Y≤(X+1)\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$1\le \text{X}\le 3 \ \text{and} \ 2\le \text{Y}\le (\text{X}+1)$$\end{document}. In the ultimate phase, we extract the secret bits from the Stego-image. The extracted secret bits helps us to regenerate the secret image which was embedded earlier. The experimental result confirms that the proposed technique offers variable payload and acceptable visual clarity. The comparison results also ensure that the proposed method is superior to the conventional edge detection based Steganography schemes as far as the payload is concerned.