In this paper, a facial feature matched one-bit transform is proposed to defect facial features and at the same time the facial orientation for face recognition. First of all the face region is segmented based on skin color. The segmented face image is then converted to a binary image with a novel facial feature matched one-bit transform that uses a kernel that is obtained by least squares. At first, eye and mouth clusters are obtained using a k-means based approach in predefined regions of the constructed binary image, and the image of the face is obtained. Then, feature vectors are extracted using discrete wavelet transform and discrete cosine transform and classification is achieved using support vector machines.