In this paper, we present a face location system in a complex background and robust to a wide range of lighting conditions, likely to appear in an indoor environment. The complete system contains two parts: the face locator and the face tracker. We will only describe the face locator. Face hypothesis are obtained combining a pyramidal greylevel template matching and a geometrical measure based on the facial feature organization. Classification to face or non-face is realized by linear discriminant analysis (LDA) on the principal components analysis (PCA) of a 26-dimensional feature vector extracted from the face hypothesis. Experiments on 1500 images in a cluttered background with 12 lighting conditions are very encouraging.