Combining Visual Saliency and Pattern Masking for Image Steganography

Recent image steganographic approaches considers perceptual characteristics of the human visual system (HVS), and often use edge detection to express them. In practice, there exist perceptually salient regions for HVS in an image according to visual attention. Moreover, HVS has different sensitivities to luminance contrast and structure of the image, i.e. pattern masking. To enhance the security of data embedding, we propose an image steganographic method in this paper which combines visual saliency and pattern masking effects. First, we perform saliency estimation and sensitivity detection from the input image. Then, we select proper thresholds and regions to embed secret data. Finally, we use least-significant-bit matching revisited (LSBMR) to embed the data, and obtain the stego image. Experimental results demonstrate that the proposed method successfully improves both image quality and embedding capacity, and is robust against visual and statistical attacks.