Face detection has been one of the most studied topics in computer vision literature; so many algorithms have been developed with different approaches to overcome some detection problems such as occlusion, illumination condition, scale, among others. Histograms of Oriented Gradients are an effective descriptor for object recognition and detection. These descriptors are powerful to detect faces with occlusions, pose and illumination changes because they are extracted in a regular grid. We calculate and vector quantizes into different codewords each descriptor and then we construct histograms of this codeword distribution that represent the face image. Finally, a set of experiments are presented to analyze the performance of this method.