BDC-GAN: Bidirectional Conversion Between Computer-Generated and Natural Facial Images for Anti-Forensics
BDC-GAN: Bidirectional Conversion Between Computer-Generated and Natural Facial Images for Anti-Forensics
复制标题
BDC-GAN：计算机生成的面部图像和自然面部图像之间的双向转换，用于反取证
DOI：
10.1109/tcsvt.2022.3177238
复制
发表时间：
2022-10
影响因子：
8.4
通讯作者：
Min Long
中科院分区：
文献类型：
--
作者：
Fei Peng;Liping Yin;Min Long
针对现有取证方法在区分计算机生成人脸图像和自然人脸图像方面存在的不足，提出了一种基于生成对抗网络的人脸图像与自然人脸图像双向转换方法（BDC-GAN）。BDC-GAN的生成器由噪声编码和内容编码两部分组成。在噪声编码中，首先利用三个高通滤波器提取图像的传感器模式噪声，然后组合堆叠的卷积层继续编码。在内容编码中，VGG-19被截断和微调以编码图像的内容。在解码器中使用了一些堆叠的卷积层和自适应实例归一化层。该算法采用多尺度图像插值。此外，对内容丢失和噪声丢失算法进行了合理的设计，并合理设置超参数，在保持原始人脸轮廓的同时实现了两个域图像之间的双向转换。实验结果和分析表明，与现有的单向CG人脸图像反取证方法和双向域自适应方法相比，该方法具有更好的视觉质量和更强的欺骗能力，并通过对现有9种取证方法的测试验证了其有效性.它揭示了现有的取证技术可以通过使用对抗学习来绕过，并且它最终将推动计算机生成和自然面部图像的区分性能的提高。
Aiming at degrading the capability of the existing forensic methods in discriminating computer generated and natural facial images, a bidirectional conversion between computer-generated and natural facial images based on generative adversarial network (BDC-GAN) is proposed for anti-forensics in this paper. The generator of BDC-GAN is composed of noise encoding and content encoding. In the noise encoding, three high-pass filters are first utilized to extract the sensor pattern noise of the image, and then the stacked convolution layer is combined to continue encoding. In the content encoding, VGG-19 is truncated and fine-tuned to encode the content of the image. Some stacked convolution layers and adaptive instance normalization layer are used in the decoder. The discriminator uses multi-scale image discriminator. Furthermore, content loss and noise loss are well designed, and hyperparameters are reasonably set to accomplish the bidirectional conversion between two domain images meanwhile retaining the original facial contour. Experimental results and analysis demonstrate that the proposed anti-forensic method can achieve better visual quality and stronger deception ability compared with the existing unidirectional CG facial image anti-forensic methods and bidirectional domain adaptive methods, and its effectiveness is verified by the tests on the existing 9 forensic methods. It reveals that the existing forensic techniques can be bypassed by using adversarial learning, and it will eventually push the performance improvement of the discrimination of computer generated and natural facial images.
登录
查看更多内容
DOI：
10.1109/icme.2018.8486579
发表时间：
2018-04
期刊：
2018 IEEE International Conference on Multimedia and Expo (ICME)
影响因子：
--
作者：
H. Nguyen;Ngoc-Dung T. Tieu;Hoang-Quoc Nguyen-Son;J. Yamagishi;I. Echizen
通讯作者：
H. Nguyen;Ngoc-Dung T. Tieu;Hoang-Quoc Nguyen-Son;J. Yamagishi;I. Echizen
基于多重分形和回归分析的自然图像和计算机生成图形的区分
DOI：
10.1016/j.aeue.2016.11.009
发表时间：
2017-01-01
影响因子：
3.2
作者：
Peng, Fei;Zhou, Die-lan;Sun, Xing-ming
通讯作者：
Sun, Xing-ming
DOI：
10.1109/icme.2007.4284852
发表时间：
2007-07
期刊：
2007 IEEE International Conference on Multimedia and Expo
影响因子：
--
作者：
Wen Chen;Y. Shi;Guorong Xuan
通讯作者：
Wen Chen;Y. Shi;Guorong Xuan
影响因子：
5.6
作者：
Yongzhen Ke;Weidong Min;Xiuping Du;Zhen-wen Chen
通讯作者：
Yongzhen Ke;Weidong Min;Xiuping Du;Zhen-wen Chen
中国
DOI：
10.1017/s0021911800113920
发表时间：
1961
期刊：
The Journal of Asian Studies
影响因子：
--
作者：
Tineke d’haeseleer
通讯作者：
Tineke d’haeseleer