In this paper a method is proposed that is meant to solve one of the subtasks of facial recognition. One of the main problems in facial recognition is trying to recognize a face from different angles, with an insufficient training set. To solve this problem it is necessary to adapt an existing three-dimensional model of the head based on two facial images (full face and profile). In this paper a semi-automatic method is proposed, that is based on modified Viola-Jones algorithm. The proposed method is used to process existing images of a person, extract the face region and main facial features from images (eyes, nose and mouth) and semi-automatically place control points on two facial images that will be later used in adapting an existing three-dimensional model of the face. The constructed model allows increasing the training set and acquiring additional information about the shape of the face.