A Robust Facial Feature Points Location Algorithm under Variable Illumination

In this paper, a novel algorithm is proposed to locate facial feature points robustly and accurately. It uses the coarse-to-fine strategy to search feature points. Firstly, it used phase congruency to model the feature points locally and locate facial feature points roughly. Secondly, it used reflectance, which is intrinsic texture of object, to drive the search of fine location. Experiments showed that our algorithm could locate the feature points more accurately and robustly, and work efficiently even under a wide range of illumination and poor initialization