LAUSR

Face recognition (FR) based on image set is an important topic in computer vision. There are numerous approaches that apply pose estimation method for single image face recognition, but few embed pose estimation method into image set-based face recognition. The conventional pose estimation method PnP used for single image needs to be modified to be fit for set-based recognition task. This study presents a method to estimate the poses of image set by applying nonlinear least squares to facial landmarks. Moreover, the distances between every single image in the query set and the ones with similar corresponding poses in the gallery sets are compared. We improve the conventional PnP method by identifying a frontal image of each image set instead of using a fixed 3-D model. Our method is evaluated on the benchmark Honda/UCSD database and YouTube Celebrities database. Experimental results show that our method leads the performance in FR based on image sets compared with other published methods.