Performance Analysis of Pose Invariant Face Recognition Approaches in Unconstrained Environments

2. Related Work

The non-intrusive nature of the face biometric makes it superior to other biometrics. Therefore, face is one of the most suitable biometrics for surveillance applications. However, in typical surveillance scenarios, people usually walk free, and it is not always possible to capture the frontal faces. This leads to a problem in face recognition, unconstrained face recognition. Although there have been some early successes in automatic face recognition, it is still far from being completely solved, especially in uncontrolled environments. Figure 1 addresses the difficult problem of identifying a face taken in uncontrolled situations with different view point using a large dictionary of still face images of thousand people. Among this, pose variations can be considered as one of the most important and challenging problems in face recognition. As the viewpoint varies, the 2D facial appearance will change because the human head has a complex non planar geometry.

Figure 1.

Overall block diagram of face recognition methodology