Zobraziť v EDS

Fourier-Based Frequency Space Disentanglement and Augmentation for Generalizable Face Anti-Spoofing.

Uložené v:
Podrobná bibliografia
Názov: Fourier-Based Frequency Space Disentanglement and Augmentation for Generalizable Face Anti-Spoofing.
Autori: Yu Y, Du Z, Luo H, Xiao C, Hu J
Zdroj: IEEE journal of biomedical and health informatics [IEEE J Biomed Health Inform] 2025 Aug; Vol. 29 (8), pp. 5413-5423.
Spôsob vydávania: Journal Article
Jazyk: English
Informácie o časopise: Publisher: Institute of Electrical and Electronics Engineers Country of Publication: United States NLM ID: 101604520 Publication Model: Print Cited Medium: Internet ISSN: 2168-2208 (Electronic) Linking ISSN: 21682194 NLM ISO Abbreviation: IEEE J Biomed Health Inform Subsets: MEDLINE
Imprint Name(s): Original Publication: New York, NY : Institute of Electrical and Electronics Engineers, 2013-
Výrazy zo slovníka MeSH: Fourier Analysis* , Face*/diagnostic imaging , Face*/anatomy & histology , Image Processing, Computer-Assisted*/methods, Humans ; Algorithms ; Female ; Male
Abstrakt: Generalizing face anti-spoofing (FAS) models to unseen distributions is challenging due to domain shifts. Previous domain generalization (DG) based FAS methods focus on learning invariant features across domains in the spatial space, which may be ineffective in detecting subtle spoof patterns. In this paper, we propose a novel approach called Frequency Space Disentanglement and Augmentation (FSDA) for generalizable FAS. Specifically, we leverage Fourier transformation to analyze face images in the frequency space, where the amplitude spectrum captures low-level texture information that forms distinct visual appearances, and the phase spectrum corresponds to the content information. We hypothesize that the liveness of a face is more related to these low-level patterns rather than high-level content information. To locate spoof traces, we disentangle the amplitude spectrum into domain-related and spoof-related components using either empirical or learnable strategies. We then propose a frequency space augmentation technique that mixes the disentangled components of two images to synthesize new variations. By imposing a distillation loss and a consistency loss on the augmented samples, our model learns to capture spoof patterns that are robust to both domain and spoof type variations. Extensive experiments on four FAS datasets demonstrate the superiority of our method in improving the generalization ability of FAS models in various unseen scenarios.
Entry Date(s): Date Created: 20240624 Date Completed: 20250806 Latest Revision: 20250807
Update Code: 20250807
DOI: 10.1109/JBHI.2024.3417404
PMID: 38913519
Databáza: MEDLINE
Popis
Abstrakt:Generalizing face anti-spoofing (FAS) models to unseen distributions is challenging due to domain shifts. Previous domain generalization (DG) based FAS methods focus on learning invariant features across domains in the spatial space, which may be ineffective in detecting subtle spoof patterns. In this paper, we propose a novel approach called Frequency Space Disentanglement and Augmentation (FSDA) for generalizable FAS. Specifically, we leverage Fourier transformation to analyze face images in the frequency space, where the amplitude spectrum captures low-level texture information that forms distinct visual appearances, and the phase spectrum corresponds to the content information. We hypothesize that the liveness of a face is more related to these low-level patterns rather than high-level content information. To locate spoof traces, we disentangle the amplitude spectrum into domain-related and spoof-related components using either empirical or learnable strategies. We then propose a frequency space augmentation technique that mixes the disentangled components of two images to synthesize new variations. By imposing a distillation loss and a consistency loss on the augmented samples, our model learns to capture spoof patterns that are robust to both domain and spoof type variations. Extensive experiments on four FAS datasets demonstrate the superiority of our method in improving the generalization ability of FAS models in various unseen scenarios.
ISSN:2168-2208
DOI:10.1109/JBHI.2024.3417404