Head pose estimation for recognizing face images using collaborative representation based classification
Srija Chowdhury, Jaya Sil · 2016
Real time face recognition is challenging due to time taken for searching the test image within a wide variation of training images. We propose an efficient face recognition method by applying collaborative representation based classification (CRC) technique in two steps. Using CRC first we select the images closer to the pose of the test image compare to others in the training set. The test image is then searched among the selected training images only, instead of all, thus reducing the search space and time. For person recognition we calculate Gabor wavelets of the eye regions of the test and the selected training images as the basis functions containing detail edge information. The images are reconstructed as a linear span of the basis vectors using CRC and the sparse coefficient vectors are used as feature vectors. To obtain the best match image we apply 12norm on the feature vectors corresponding to the test and the selected images. The proposed head pose estimation based face recognition method has been validated using PIE and Head Pose databases and obtain comparable accuracy with other methods at a much lesser storage and time complexity.