Head Pose-Based Conditional Regression Forest for Facial Feature Detection

Multi-angles of facial feature detection is still a challenging research. In this paper, the author proposes a precision head pose estimation method as a condition to improve the performance of regression forests, and decreases the missing rate caused by head deflection. The basic idea is used by locality preserving projection, a kind of manifold learning, and nonlinear regression (LPP+NLR) for getting the global information of pose and label it, then utilize trained conditional regression classifier to identify the feature points in global characteristics. The effectiveness of the proposed facial feature detection algorithm is illustrated in the experiments and the comparison with several recent methods.