Image-projection-based facial feature vector design

Selecting facial feature regions of interest (ROIs), together with the ensuing image analysis, allows for the extraction of diverse facial feature information, and is a common strategy for face recognition. However, the traditional facial feature ROIs must often be hand-selected or automatically detected using prior knowledge. This study proposes to project pixel information from the eyes and mouth onto the horizontal and vertical axis to generate a facial feature pixel statistics graph that can be used to analyze eye and mouth ROIs. Image analysis within these ROIs gives the positions of the left and right eye corners, which can be used as facial feature vectors, and treated as input to a support vector machine (SVM). The experimental results of this study suggest that the proposed eye and mouth selection and facial feature vector facilitates face recognition.