Abstract :
The aim of our work is automatic facial expression analysis based on the study of temporal evolution of facial feature boundaries. Previously, we developed a robust and fast algorithm for accurate lip contour segmentation (Eveno, N. et al., IEEE Trans. Circuits and Systems for Video Technology, 2004). Now, we focus on eye and eyebrow boundary extraction. The segmentation of eyes and eyebrows involves three steps: first, an accurate model based on flexible curves is defined for each feature; second, models are initialized on the image to be processed after the detection of characteristic points such as eye corners; third, models are accurately fitted to the facial features of an image according to some information of luminance gradient. The performance of our method is evaluated by a quantitative comparison with a manual ground truth and also by the analysis of expression skeletons based on the results of our facial features segmentation.
Keywords :
brightness; face recognition; feature extraction; gesture recognition; image segmentation; object detection; automatic facial expression analysis; automatic segmentation; characteristic point detection; expression skeletons; eye parametric model; eyebrow parametric model; facial feature boundary extraction; facial features segmentation; luminance gradient; manual ground truth; Circuits and systems; Eyebrows; Eyes; Face detection; Facial features; Image segmentation; Parametric statistics; Performance analysis; Robustness; Skeleton;
