| Presentation | 2022-03-08 [Poster Presentation] Study on JPEG Compression Resistant Watermarking Method Trained with Quantized Activation Function Shingo Yamauchi, Masaki Kawamura, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | We propose a watermarking method that introduces a quantized activation function to acquire robustness against quantization for JPEG compression. Neural network-based methods have been proposed as watermarking methods that are robust against various types of attacks. Zhu et al. proposed a watermarking method that is robust against JPEG compression, clipping and Gaussian blurring. They showed that it was possible to increase the robustness by including an additional layer that simulated these attacks between the generation layer of the stego image and the extraction layer of the watermarks. However, the method of Zhu et al. did not provide sufficient robustness against JPEG compression compared to other attacks. This was because their method did not implement the quantization function in JPEG compression. Therefore, we propose a quantized activation function, which consists of shifted hyperbolic tangent functions. In this study, we introduce the quantized activation function to the watermarking method proposed by Hamamoto and Kawamura, and evaluate the effect of quantization. Their method has no attack layer, by introducing an attack layer with a quantized activation function, we can evaluate the JPEG compression robustness. The bit error rate (BER) was used to evaluate the robustness, and the PSNR was used for image quality. As a result, although the image quality decreased by introducing the quantized activation function, the watermarks could be extracted with the BER of less than 0.1 at the Q-value of 20 or higher in JPEG compression. Therefore, by introducing the quantized activation function, the compression robustness was achieved. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | digital watermarking / autoencoder / JPEG compression / activation function |
| Paper # | EMM2021-110 |
| Date of Issue | 2022-02-28 (EMM) |
| Conference Information | |
| Committee | EMM |
|---|---|
| Conference Date | 2022/3/7(2days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | (Primary: Online, Secondary: On-site) |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | Image and Sound Quality, Metrics for Perception and Recognition, Human Auditory and Visual System, etc. |
| Chair | Ryoichi Nishimura(NICT) |
| Vice Chair | Masaaki Fujiyoshi(Tokyo Metropolitan Univ.) / Masatsugu Ichino(Univ. of Electro-Comm.) |
| Secretary | Masaaki Fujiyoshi(Utsunomiya Univ.) / Masatsugu Ichino(NICT) |
| Assistant | Shoko Imaizumi(Chiba Univ.) / Youichi Takashima(Kaishi Professional Univ.) |
| Paper Information | |
| Registration To | Technical Committee on Enriched MultiMedia |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | [Poster Presentation] Study on JPEG Compression Resistant Watermarking Method Trained with Quantized Activation Function |
| Sub Title (in English) | |
| Keyword(1) | digital watermarking |
| Keyword(2) | autoencoder |
| Keyword(3) | JPEG compression |
| Keyword(4) | activation function |
| 1st Author's Name | Shingo Yamauchi |
| 1st Author's Affiliation | Yamaguchi University(Yamaguchi Univ.) |
| 2nd Author's Name | Masaki Kawamura |
| 2nd Author's Affiliation | Yamaguchi University(Yamaguchi Univ.) |
| Date | 2022-03-08 |
| Paper # | EMM2021-110 |
| Volume (vol) | vol.121 |
| Number (no) | EMM-417 |
| Page | pp.pp.95-100(EMM), |
| #Pages | 6 |
| Date of Issue | 2022-02-28 (EMM) |