Face detection in images or video sequences is a very challenging problem. It has a wide range of applications but at the same time it presents a great number of difficulties, since faces are non-rigid and very changeable objects that can adopt a lot of different poses and with a high inter and intra-person variation and a high sensitivity to lighting conditions. Along this document, a new approach to the face detection and pose estimation problem is given. This approach is based on the method proposed by Viola and Jones in  but considering a wide range of face poses, varying the elevation and the out-of-plane rotation, and building specific classifiers for each one. The proposed method can be easily adapted to consider other poses or to detect other objects. Especially, this approach is interesting when an object that can adopt several positions want to be detected, since the partition of the pose space allows to build classifiers specialised in only one or a few poses, which limits the large variance of the global class, the class containing all the poses. In order to facilitate the reproduction of all the processes done in this document, we have used standard face datasets to train and test the system.