Face Anti-spoofing Detection Based on Novel Encoder Convolutional Neural Network and Texture’s Grayscale Structural Information
جاري التحميل...
التاريخ
عنوان الدورية
ردمد الدورية
عنوان المجلد
الناشر
International Journal of Computational Intelligence Systems, Springe
خلاصة
The rise in popularity of face anti-spoofing is attributed to its crucial role in safeguarding face recognition systems.
Perpetrators use either a photo or a video to execute a face-spoofing attack on the authentication system of an
authorized user, aiming to gain access to the user’s resources. Traditional face anti-spoofing detection techniques
often rely on grey texture elements while disregarding RGB color intensity features in face images, which lead to
a loss of certain facial information. On the other hand, RGB-based methods could introduce noise or superfluous
elements that reduce the effectiveness of spoofing detection. Furthermore, in real-world situations, color disparities may be introduced by lighting, camera settings, and other elements. Therefore, a face anti-spoofing detection
approach based on a novel Encoder Convolutional Neural Network (ECNN) architecture, Local Binary Pattern
(LBP) and Local Ternary Pattern (LTP) descriptors is proposed. The proposed ECNN can more effectively extract
the RGB color brightness features. It encodes the brightness intensity change of the RGB color transition, which
can recognize intricate spoofing attempts. The texture descriptors of LBP and LTP are utilized to extract the texture's grayscale structural information to overcome different environmental situations. This work combined the
human face's intensity RGB color and grey texture categorization parameters to increase the detection accuracy
of face spoofing and produce positive experimental outcomes. We contrast our proposed method with alternative
algorithms and verify its performance using three publicly available datasets, showcasing its superiority.