Data Embedding Method for Printed Images using Deep Neural Networks
Hideaki Orii, Takaharu Kouda, Hideaki Kawano · 2019
Data embedding techniques for images, also known as watermarking and steganography, are used in various applications such as copyright protection and security. In recent years, data embedding methods for images is applied not only to electronic media but also to printed textures and other patterns for information presentation. In some of them, a digital watermarking method using DCT (Discrete Cosine Transform) has been used. In this paper, we propose a novel data embedding method for printed images using deep neural networks. In the proposed method, the process of embedding and restoring data is represented by one deep neural network, and the embedding process into DCT coefficients is optimized by learning of the deep neural network. In the experiment, we apply the proposed method to the printed images have various textures. The results showed the effectiveness of proposed method.