In biometric identification, face recognition is one of the most active research areas. Different face recognition models proposed with the acceptable performance mostly under the supervised conditions. However, now days face recognition under uncontrolled diseases such as low-resolution images, internet downloaded images, mobile and surveillance captured is gained significant researchers attention. The significant variations in face illumination, pose, occlusion, and image quality are key state-of-art challenges for the robust face recognition. In this paper, we proposed a novel face recognition model to address the challenges of face images collected under constrained and unconstrained conditions. For the face representation, we proposed Hybrid Dual Cross Pattern (H-DCP) to solve the low-resolution face images recognition problem efficiently. First, we exploit laplacian filter to lower the illumination variations impact and then extract H-DCP features at the component and holistic levels. DCP based face descriptor method is efficient regarding feature extraction time as well as the accuracy of recognition. The face matching is done by using Feed Forward Neural Network (FFNN) using Principal component analysis (PCA) called PFFNN. PFFNN classifier is designed and trained with different face databases to evaluate the performance of proposed work. The simulation results on three well-known datasets (LFW, CK, and CAS) claims robustness against the state-of-art methods.
Volume 11 | 01-Special Issue
Pages: 708-719