Local adjacent extrema pattern for fingerprint image classification

Abstract

Fingerprint classification is as yet difficult issue due to incomplete information in ridges and low quality images. To address these problems, we propose a new and robust Local Adjacent Extrema Pattern (LAEP) is a feature descriptor for fingerprint classification. The proposed descriptor finds it values based on indexes of local adjacent extremas using first order derivatives. The intensity values of the local extremas are compared with center pixel intensity value to employ the correlation of central pixel with its neighbors. Finally, the descriptor is generated on the support of the indexes and local extremas values. Support Vector Machine (SVM) is utilized for classification of fingerprint images into five classes. To prove the effectiveness of the proposed descriptor, we have tested on Indraprastha Institute of Information Technology (IIIT) - Indian rural fingerprint database for classification. In addition, the classification results of the proposed descriptor are compared with the existing methods. The resultant LAEP descriptor proved better classification accuracy than the previous methods.