Face recognition has attracted researches not only because faces represent a challenging class of naturally textured 3-D objects, but also because of many applications of automatic face recognition, such as: physical security, crowd surveillance, face reconstruction, and many more. In this paper, an image-based face recognition model is proposed, that verifies human faces by projecting face images onto a feature space that spans the significant variations among known face images, by computing their significant features: Discrete Cosine Transform (DCT). They do not necessarily correspond to features such as eyes, mouth, etc. These “DCT coefficients" composes the feature vector that represents a human face. An average of 98.5% recognition rate is reached on the Olivetti Research Laboratory (ORL) database of facial images [8].