An Improved Algorithm for Frontal Face Feature Location Based on ASM

In this paper, we propose an improved algorithm for frontal face feature location based on Active Shape Models (ASM). There are two improvements in this algorithm. Firstly, to improve the accuracy of the initialization about average face model, we initialize the model using the size of eye and the location of the face in facial image. Secondly, to improve the accuracy and robustness of matching model points to target points in the algorithm, multi-level bilateral filtering is implemented. According to the times of filtering, the facial image is divided into different levels. This involves searching for the object from the least smoothing image that in the bottom level. Then points´ location is refined in more smoothing image.