Extraction réaliste des traits caractéristiques du visage à l’aide de modèles paramétriques adaptés

Abstract

In this paper, we are dealing with the problem of facial features segmentation (mouth, eyes and eyebrows). A specific parametric model is defined for each feature, each model being able to take into account all the possible deformations. In order to initialize each model, some characteristic points are extracted on each image to be process (for example, the corners of the eyes, mouth and eyebrows). In order to fit the model with the contours to be extracted, a gradient flow (luminance or chrominance ) through the estimated contour is maximized because at each point of the searched contour, the gradient (luminance or chrominance) is normal. The advantage of the definition of a model associated to each feature is to be able to introduce a regularisation constraint. However, the chosen models are flexible enough in order to procude realistic contours for the mouth, the eyes and eyebrows. This facial features segmentation is the first step of a global system of facial emotions recognition.