Palm Print Recognition using Modified Local Features based on Sparse Classifier

Abstract

In this paper, a novel algorithm for palm print recognition using combination of image features in coarse and fine scales is introduced. The algorithm is implemented on the PolyU databases using MATLAB. In the preprocessing stage of this algorithm, speeded-up robust features (SURF) method is applied on the captured image, to tackle scaling, rotation and translation problems. In the feature extraction stage, two coarse and fine scale features are extracted and combined. In the coarse scale, a general categorization using shape context descriptor is done. Then, Histogram of Oriented Gradient features are extracted from the image, in the fine scale. In the classification stage, sparse coding is used for classification feature matrices in order to obtain a correct recognition rate. The features used in this paper are robust against changes of the scale, rotation and illumination. The recognition rate of the proposed approach is 99.85%, approximately.