Image Hide with Invertible Network and Swin Transformer

Image hiding is a way of hiding information by hiding a secret image in a carrier image in an imperceptible way and recovering it. How to effect better hiding of images in images is a problem that is still being studied. In this paper, we propose an invertible neural network based model using the Swin Transformer module to hide images. According to the properties of invertible neural networks, image hiding and revealing can be done by the same network of forward and backward processes. Since image hiding and recovery are forward and backward of the same network, sharing the same set of parameters, a lot of resources are saved accordingly. It is found that hiding secret information in wavelet domain can improve the concealment, so we transform the image to wavelet domain for hiding.