Face liveness detection with component dependent descriptor

By Jianwei Yang, Zhen Lei, Shengcai Liao, S. Li. 2013

280 citations recorded by the source corpus.

Spoofing attacks mainly include printing artifacts, electronic screens and ultra-realistic face masks or models. In this paper, we propose a component-based face coding approach for liveness detection. The proposed method consists of four steps: (1) locating the components of face; (2) coding the low-level features respectively for all the components; (3) deriving the high-level face representation by pooling the codes with weights derived from Fisher criterion; (4) concatenating the histograms from all components into a classifier for identification. The proposed framework makes good use of micro differences between genuine faces and fake faces. Meanwhile, the inherent appearance differences among different components are retained. Extensive experiments on three published standard databases demonstrate that the method can achieve the best liveness detection performance in three databases.

Subfields: Computer Vision and Pattern Recognition, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Signal Processing

Citations in the corpus

Biometric Antispoofing Methods: A Survey in Face Recognition (2014, methodology): “combination of low-level and high-level face component descriptors [146]; a recent trend based on the use of Local Binary Patterns (LBP) to detect photo-attacks [104], [147], which has been successfully combined with other texture descriptors such as Gabor Wavelets and with shape related information”
Deep Spatial Gradient and Temporal Depth Learning for Face Anti-Spoofing (2020, methodology): “The early works usually rely on hand-crafted features, such as LBP [14, 15, 37], SIFT [42], SURF [9], HoG [28, 52], DoG [43, 48], and traditional classifiers, such as SVM and Random Forests.”
Face anti-spoofing using patch and depth-based CNNs (2017, methodology): “Common local features that have been used in prior work include LBP [30, 15, 16], HOG [25, 47], DoG [42, 36], SIFT [34] and SURF [9].”
Learn Convolutional Neural Network for Face Anti-Spoofing (2014, background): “These improvements suggests that background region is indeed helpful for face anti-spooﬁng to some extent when using CNN-learned or hand-crafted features [35].”
Learning Deep Models for Face Anti-Spoofing: Binary or Auxiliary Supervision (2018, methodology): “Researchers start the texture-based anti-spooﬁng approaches by feeding handcrafted features to binary classiﬁers [13,18, 19,27,33,34,38,49].”
Deep Tree Learning for Zero-Shot Face Anti-Spoofing (2019, background): “, LBP [7,16,17,35], HoG [25,47], SIFT [38] and SURF [8], with traditional classifiers, e.”
Face Anti-Spoofing: Model Matters, so Does Data (2019, methodology): “For example, many hand-crafted features have been studied in previous works, including LBP [29, 12, 13], HOG [21, 43], DOG [37, 33], SIFT [32], and SURF [9].”
CASIA-SURF: A Large-Scale Multi-Modal Benchmark for Face Anti-Spoofing (2019, methodology): “All of these methods use handcrafted features, such as LBP [34]–[37], HoG [36]– [38] and GLCM [38].”
Secure Face Unlock: Spoof Detection on Smartphones (2016, background): “Face texture analysis [8]–[12] Relatively low computational cost and fast response Poor generalizability, Requires face and/or landmark detection Idiap Replay-attack (Intra-DB, Cross-DB) [11]: (15.”
Face Spoofing Detection Using Colour Texture Analysis (2016, background): “However, these are shown to be important factors already in intradatabase tests [6], [10], [56] and we noticed that they also significantly affect the cross-database performance.”
CelebA-Spoof: Large-Scale Face Anti-Spoofing Dataset with Rich Annotations (2020, background): “Most traditional algorithms focus on handcrafted features, such as LBP [5,26,27,35], HoG [26,35,30] and SURF [2].”
Face anti-spoofing based on color texture analysis (2015, unknown): “ive results on the Replay-Attack database. Table 3. Comparison between the proposed countermeasure and state-of-the-art methods Replay-Attack CASIA Method EER % HTER% EER% IQA based [12] - - 32.4 CDD [13] - - 11.8 DOG (baseline)[10] - - 17.0 Motion+LBP [14] 4.5 5.1 - Motion [6] 11.6 11.7 26.6 LBP [4”
Face De-Spoofing: Anti-Spoofing via Noise Modeling (2018, methodology): “Early works apply the hand-crafted feature descriptors, such as LBP [3,4,21], HoG [5,6], SIFT [22] and SURF [23], to project the faces to a low-dimension embedding.”