Date of Publication :29th March 2018

Abstract: Facial Feature Detection is one of the essential techniques in face recognition, face modeling, head pose estimation and facial expression recognition. Emotions may be recognized through the facial features such as eyes, nose, lip movements etc. The objective of the proposed work is to detect the face with these facial features. Cascade Object Detector is initially proposed to identify the face region from the input image. Then Regression based face alignment algorithm is employed for the feature point alignment and registration. By means of the scale-invariant feature transform and regression results, the landmark points for the facial features are estimated inside the template. Delaunay method is used to construct a triangulation that can be utilized to draw the template with the associated points. Finally the facial features are detected along with the boundary points. The WSEFEPv101lo dataset is used for evaluation and analysis. Out of 62 facial images, the facial features of 48 images are accurately detected and due to misalignment of facial landmark points due to facial emotions, 14 images are not detected correctly

Reference :