التفاصيل البيبلوغرافية
| العنوان: |
Face forgery detection by progressively enhancing spatial and frequency-aware features. |
| المؤلفون: |
Qi, Yongfeng, Wen, Shengcong, Zhang, Hengrui, Liang, Anye, Chen, Huili, Cao, Panpan |
| المصدر: |
Multimedia Systems; Jun2024, Vol. 30 Issue 3, p1-16, 16p |
| مستخلص: |
Due to the security issues caused by face synthesis technology, face forgery detection has received considerable attention. Therefore, there is an increasing necessity to develop effective and generalized face forgery detection models. Some current methods attempt to use frequency features to reveal clues hidden under fake faces. However, the frequency information utilized by these methods is often coarse-grained perceptual features generated by frequency transformation, which makes it difficult for them to extract fine-grained forgery traces in the ordinary learning process. To compensate for these shortcomings, we propose a learning framework that progressively enhances spatial and frequency-aware features, consisting of three carefully designed modules. Specifically, the first is the fine-grained frequency-aware extraction module, which decomposes the RGB image into fine-grained components and extracts content-related frequency information through multiple learnable filters to obtain the frequency-transformed fine-grained perceptual features, which provide fine-grained forgery information for subsequent learning of the network. The second is the noise residual enhancement module, which enhances and guides the network to capture forgery traces from the perspective of image noise. The last is the dual-domain feature attention module, which learns and enhances forgery clues in each other’s features by fusing information from RGB and multi-frequency fine-grained perceptual features. Extensive experimental results and visualizations demonstrate that our proposed model outperforms previous face forgery detection models. [ABSTRACT FROM AUTHOR] |
|
Copyright of Multimedia Systems is the property of Springer Nature and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.) |
| قاعدة البيانات: |
Complementary Index |
