Minimal distortion steganography using well-defined functions

The present paper proposes a novel method for grayscale image steganography which provides minimal distortion. The cover image is first divided into some blocks such that each block contains three pixels. Three bits of secret data are then embedded into each block by increasing/decreasing one pixel or rarely three pixels by one as done in LSB matching. However, data bits are not directly extracted from the LSBs of the pixels, but they will be obtained from three well-defined binary functions. Using the novel proposed technique, each pixel of the cover image is modified with the probability of 0.375 on average. The experimental results evaluated on 2000 natural images with RS steganalytic algorithm show that our proposed method can significantly decrease the image distortion and enhance the security.