Light Field Messaging With Deep Photographic Steganography

A Compact Neural Network-based Algorithm for Robust Image Watermarking

“…After that, RedMark [49] uses two Fully Convolutional Neural Networks (FCNs) with residual connections to embed watermarks in the frequency domain without adversarial training. Different from the dependent deep hiding methods (DDH) [8], [14]- [17], which adapt the watermark to the original cover image, UDH [50] proposes a universal deep hiding method to embed the watermark independent of the cover image. These existing works have demonstrated a variety of neural network structures to effectively realize the message embedding and extraction, ensuring that the watermarked image and the cover image have little or even no perceptual differences.…”

“…For non-differentiable distortions including JPEG compression, some methods [7], [8], [49] turn to simulate them with a differentiable approximation, allowing the network to be trained in an end-to-end style. In [9] and [14], some distortions are generated by a trained CNN instead of explicitly modeling distortions from a fixed pool during training, which is another way to deal with non-differentiable and hard modeled distortions. In addition, Liu et al [51] design a redundant two-stage separable deep learning framework to address the problems in one-stage end-to-end training, such as image quality degradation and difficulty to simulate noise attacks using differentiable layers.…”

“…Besides the noise layer, most existing robust image watermarking systems based on deep learning use Encoder-Noiser-Decoder frameworks [7], [8], [14]- [17]. Among them, the encoder embeds the watermark into the cover image in an imperceptible manner, and the decoder recovers the watermark message from the distorted watermarked image.…”

“…In addition, neural networks can be retrained to resist novel types of attacks, or to focus on particular features such as the robustness and imperceptibility without designing a new specialized sophisticated algorithm, enabling them possible to develop an adaptable and generalized framework for various watermarking applications [7]. However, most of them use the Encoder-Noiser-Decoder framework [7], [8], [14]- [17], as shown in Fig. 1 (a).…”

DeepFormableTag

“…HiDDeN [Zhu et al 2018] augments messages by applying color distortions and noise to the encoded image, assuming perfect alignment during the decoding process without any spatial transformations. Wengrowski and Dana [2019] capture a dataset to model a camera-display transfer function in order to add distortions to input images. However, none of these methods account for the localization problem of encoded messages in input images.…”