Face detector locates all the faces in a given image. A size varying window scans the given input image and passes them to face detector, which classifies them info face or non-face. Instead of finding face, as a whole(holistic approach), smaller facial components could be detected and post-processed for determination of presence of face. Detected facial components such as eye, mouth and nose are used to construct a geometric graph which is matched with reference model. A detector gives location, size, type and confidence of its detection. In this thesis, confidence is used for false positive elimination. And it is also used while calculating the cost of matching with the reference model. A graph with lowest cost and below threshold is chosen as face. This component based method is more robust in varying illumination, pose and partial occlusion.