A Literature Survey of Face Recognition Under Different Occlusion Conditions
Manisha Kumari Meena, Hemant Kumar Meena · 2022 IEEE Region 10 Symposium (TENSYMP) · 2022
Face recognition is a very old and popular biometric problem. In the last two decades, several challenges occur in face recognition such as expression, occlusion, illumination, pose variation, and low-quality image. Also, several techniques exist, such as Principle Component Analysis (PCA), Linear Discriminate Analysis (LDA), Local Non-Negative Matrix Factorization (LNME), Independent Component Analysis (ICA), and other variants, which finds good results under constrained scenarios. But, they give poor accuracy in an unconstrained scenarios such as illumination, pose variation, and occlusion. As Face recognition biometrics has gained great importance in occlusion condition, for example, a terrorists cover his face using facial accessories like sunglasses, scarves, or masks. This occlusion face recognition is categorized into two parts. 1) Feature extraction without occlusion detection face recognition. 2) Feature extraction with occlusion detection face recognition 3) occlusion recovery face recognition. Feature-based and fractal-based methods consider features around the eyes, nose, or mouth region to be used in the recognition face of the algorithm. This paper describes the different techniques and databases to handle the partial occlusion problem.