A high speed method for features extraction in face recognition systems

Face recognition is one of the most important and practical issues in image processing, and descriptors are used in order to extract image features. These descriptors are used in the face recognition scheme in terms of speed, processing time, and the extracted feature vector dimensions. The method presented in this paper uses local difference binary descriptor and local quantized patterns, which is used in the local difference binary descriptor from the features of average intensity of the pixels, horizontal and vertical gradients and in the local quantized pattern, from the Fourier transform. Because the face recognition scheme has a large database, it is necessary to classify the extracted features using the support vector machine classifier and the nearest neighbor k. The ORL database was used to evaluate the proposed method, which achieved the average recognition rate 97 percentage and total processing time has been at least 80% faster. The evaluation results show the proposed method's proper performance in rapid face recognition scheme and real-time face recognition schemes.