An end-to-end image hiding model based on skip connected dense block and edge loss

End-to-end image hiding is a new hiding paradigm which hides an image into another image. However, the existing end-to-end image hiding models still do not have a good balance between imperceptibility and computational complexity. To deal with the problem, a novel end-to-end image hiding model is proposed based on skip connected dense block and edge loss in this paper. Firstly, a new type of hiding/extraction network is designed by utilizing dense block and skip connection. Secondly, edge loss is defined to measure the difference in detail between the predicted distribution and the real distribution. In a covert communication, the sender needs to input a cover image and a hidden image to the trained hiding network, and a container image which has no visual difference from the cover image will be produced. For the receiver, the hidden image can be recovered from the container image by using the extraction network. Experimental results and analysis demonstrate the effectiveness of the proposed model, compared with the existing end-to-end image hiding models, the container image and the hidden image generated by the proposed model can achieve better visual quality as PSNR 43.96 dB and SSIM 0.987, and the effective capacity is the highest 17.7458 bpp. It has broad prospects for high-capacity covert communication.