Adaptive edge tracing steganography

The present paper proposes an adaptive edge tracing method for grayscale image steganography. The regions located in the edges present more complicated statistical features and thus it is more difficult to observe changes in the edges than those in smooth regions. In the following proposed method, firstly, all edges of the cover image, both horizontal and vertical, are fully detected. To do this, we compute the gradient magnitude of the pixels of cover image using the Sobel operators. Secondly, according to the length of secret data, we adaptively preserve the sharper edges and suppress the weaker edges. For the lower length of data, the sharper edges have the higher priority to be used. When the length of data increases, more edges will be used for data embedding. Thirdly, we trace the edge pixels in a novel manner and make some groups which each group contains three consecutive pixels. Finally, we embed two bits of secret data into the LSB of the pixels in each group using matrix embedding. By this technique, we need to change at most one LSB in a group. Therefore, the embedding efficiency is improved. The experimental results evaluated on 10,000 natural images with a steganalytic algorithm show that our proposed method can significantly enhance the security and can preserve higher visual quality of the stego-images.