An Efficient Way of Hybridizing Edge Detectors Depending on Embedding Demand

Abstract

Edge detection-based image steganography schemes usually embed data in edge pixels only. However, some schemes embed data in non-edge pixels as well. In that case, the schemes embed more bits in the edges than in the smoothed areas. In all cases, the schemes perform large changes in a tiny area of the image during small data embedding. Detecting such local modifications is comparatively easier for a steganalyzer. As a result, it is preferable to distribute bits evenly across the image. Again, the schemes struggle to hide large messages in a cover image due to the algorithmic approach of hiding a fixed number of bits per pixel. In this research, we have overcome that problem by employing multiple edge detectors in generating a resultant edge image. Depending on the embedding needs, we have checked whether a single edge detector is sufficient to help in conceiving all bits or not. If it is not possible for a single-edge detector, we have hybridized them. Hybridization of edge images is performed either by logical AND, OR or OR with dilation. When the message size is very small, we have generated the resultant edge image by doing a logical AND operation among the edge images. That strategy have reduced the number of edge pixels as well as helped in distributing the to-be-embedded bits over the image in a more evenly manner. Similarly, to meet a larger embedding demand, we have performed a logical OR operation among the same edge images to increase the number of edge pixels. Even, to meet more embedding demand, we have dilated the OR-resultant image. These processes were carried out dynamically in the research according to an embedding demand. The experimental results deduce that this scheme embeds 92.37%, 73.92%, 74.78%, and 9.60% more bits than four competing methods. Similarly, for small embedding demand, the proposed scheme demonstrates 37.45%, 46.87%, 44.21%, and 55.56% higher PSNR values than the others. Moreover, statistical analyses state that this scheme demonstrates stronger security against attacks.