Efficient Gabor Feature Based Approach for Face Recognition

摘要

This paper investigates the use of geometrical facial features retrieved from normalized images for face identification, and verification of frontal and near frontal facial images. A set of feature points representing relative locations of facial features was selected. By selecting multiple feature points within the salient facial features, we intend to cover the full extent of the features. The positions of the eyes are set at preset fixed coordinates after normalization and the relative positions of nose and mouth are determined intuitively using anthropometric standards by taking distance measures to other salient features. Extracted Gabor features at respective feature locations are used to derive a measure for similarity detection. Test results of the technique and conclusions are presented.