Face Recognition Under Bad Illumination Conditions

Accurate face recognition in variable illumination environments has attracted the attention of the researchers in recent years, because there are many applications in which these systems must operate under uncontrolled lighting conditions. To this end, several face recognition algorithms have been proposed which include an image enhancement stage before performing the recognition task. However, although the image enhancement stage may improve the performance, it also increases the computational complexity of face recognition algorithms. Because this fact may limit their use in some practical applications, recently some algorithms have been developed that intend to provide enough robustness under variable illumination conditions without requiring an image enhancement stage. Among them, the local binary pattern and eigenphases-based schemes are two of the most successful ones. This paper presents an analysis of the recognition performance of these approaches under varying illumination conditions, with and without image enhancement preprocessing stages. Evaluation results show the robustness of both approaches when they are required to operate in illumination varying environments.