Face Recognition Using Multi-Resolution Transform

Face recognition has wide range potential applications in commercial and law enforcements, such as, security surveillance, telecommunication, human computer interaction. This paper deals with a novel technique of face recognition using multi-resolution transform such as, Gabor wavelet transform. Multi-scale or resolution methods are based on image transformations that analyze the image at multiple resolutions. Gabor wavelet is used to extract the spatial frequency, spatial locality and orientation selectivity from faces irrespective of the variations in the expressions, illumination and pose. Normalization is done to reduce dimensionality which will reduce memory problem and computation time. Principal component analysis (PCA) deals with the decomposition of the training set into the eigenvectors called eigen faces. Then by considering each eigen faces as each co-ordinate, a co-ordinate system is formed called face space. In this face space, each face is considered as a point. All samples in each class forms the cluster of points in the face space. By projecting each faces, its co-ordinate values can be determined, which are later used for distance measures in discrimination analysis. Various discrimination analyzes such as, Euclidean, L1, L2 and cosine similarity are used for the recognition of face images.