A Hierarchal Framework for Finger-Vein Image Classification

Abstract

For personal identification, the biometric systems based on finger-vein pattern have been successfully used in many applications. The concern for the system efficiency over a large database should not be negligible in the real situation. So, categorizing the finger-vein images to different classes is helpful for reducing pattern matching cost. In this paper, we propose a level-based framework for roughly and automatically categorizing finger-vein images. The proposed level-based framework consists of two layers in classifying finger-vein images. In this framework, the imaging qualities and the image contents are respectively used for the first layer and second layer image feature representation. And the k-means algorithm is adopted for automatic finger-vein image clustering. Using SVM scheme, we can achieve 99% CCR (correct classification rate) over a large image database. Finally, for comparison, the POC (Phase-Only-Correction) matching algorithm is used. Experimental results show that the proposed method has a good performance in the improving recognition efficiency as well as recognition accuracy.