Resumen
This paper presents a novel technique for Face Recognition from a Partial Face View (FRPV), which consists of three phases. The first phase uses an existing algorithm to detect faces in input images. The second phase includes splitting the input images undetected by the first phase into two, four, six, or eight parts. Then, every part is rotated by a new split and rotate face detection (SRFD) algorithm until it detects a face in one of these partial images. The third phase uses the Eigenfaces method with train and test databases to perform recognition. This phase compares the selected test image with images in the train database until it recognizes the person and updates the train database. The FRPV system was implemented using a head-pose image database where every person has multiple images with several poses having different Pitch and Yaw Angles ranging from ?90º to +90º. The results showed that the FRPV system outperformed previous methods. Its accuracy rate was equal to 96% for faces that had different poses. In addition, the SRFD method achieved a detection success rate of 67%, which is better than other similar methods.