This paper proposes a face detection approach,which utilizes the modified Census transform,histogram intersection kernel (HIK) support vector machine (SVM) classifier and a simplified Haar method.Compared with the face detection methods of Haar + Adboost and LBP + Adboost,the proposed appraoch can improve the detection accuracy by reducing the false positive and false negative detection.Moreover,the proposed technique shows illumination invariance.In addition,the omit of the Sobel transform in this appraoch improves the detection speed without compromising the detection accuracy.