Research On Non-Destructive Image Information Hiding Method Based On Reversible Image-to-Image Translation

Image-to-image translation refers to the conversion of one type or style of image into another type or style of image.This involves many aspects of computer vision,computer graphics,and image processing,which are important for improving people’s quality of life and creativity.However,image translation also faces many challenges,such as how to make the translated images high-quality,reversible,and secure.Reversible information hiding technology is a technique that embeds secret information while protecting the content information of the carrier from being leaked.This technology not only can extract the embedded secret information completely but also ensures that the original carrier can be restored without loss.Applying reversible information hiding to image translation can enhance the reversibility and security of image translation;applying reversible image translation to reversible information hiding can protect data privacy.The main research work and innovation points of this paper are as follows:1.Proposed an image reversible translation methodThe image-to-image translation task,which is widely studied by generative adversarial networks(GANs),aims to map images from the source domain to the target domain.However,existing GAN-based solutions lack the ability to achieve reversible translation.To address this issue,this paper proposes an approximately reversible image-to-image translation method,where the reconstructed source image has less distortion compared to the source image because this method adopts a joint interframe coding and embedding strategy.In this strategy,we compress the source image reconstructed based on CycleGAN and the pseudo video composed of source images,and then reversibly embed the compressed bitstream into the translated image to obtain a tagged translated image.This strategy can balance the image quality of both reconstructed source images and tagged translated images.The reconstructed source image can be obtained by extracting and decoding the bitstream from the tagged translated image.Experimental results show that the reconstructed source image and tagged translated image of our proposed approximately reversible image translation method have high image quality.2.Proposed a transform-domain reversible information hiding method based on image reversible translationAlong with the popularity of data outsourcing to the cloud,it has become crucial to protect data privacy while making it easy for cloud servers to manage data.In such a demand,reversible data hidden in encrypted images has attracted more and more researchers’ attention.This article proposes a reversible information hiding method based on image reversible translation in the transform domain.The previous encryptionbased ciphertext domain information hiding framework directly encrypts images into snowflake point images,which are more likely to attract cloud attention.The proposed reversible information hiding method based on image reversible translation in the transform domain can translate images into other images,which not only protects the privacy of the original image but also does not attract cloud attention.Cloud servers can also embed watermarks using reversible information hiding methods for plaintext images.This method can also meet different requirements for image quality and embedding capacity.Experimental results show that this method has high performance in terms of image quality and security.