Face Spoofing Detection Based on Multiple Descriptor Fusion Using Multiscale Dynamic Binarized Statistical Image Features

摘要: Face recognition has been the focus of attention for past couple decades and, as a result, significant progress made in this area. However, problem spoofing attacks can challenge face biometric systems practical applications. In paper, an effective countermeasure against based on kernel discriminant analysis approach is presented. Its success derives from different innovations. First, it shown that recently proposed multiscale dynamic texture descriptor binarized statistical image features three orthogonal planes (MBSIF-TOP) detecting attacks, showing promising performance compared with existing alternatives. Next, by combining MBSIF-TOP blur-tolerant descriptor, namely, local phase quantization (MLPQ-TOP) representation, robustness attack detector be further improved. The fusion information provided and MLPQ-TOP realized via fast (KDA) technique. It avoids costly eigen-analysis computations solving KDA spectral regression. experimental evaluation system databases demonstrates its advantages various imaging conditions, methods.