Texture Development Through Lossless Compression of Biometric Representative for Palmprint Recognition System

Abstract

Local binary pattern (LBP) features have been used widely in texture extraction because of its simplicity and high accuracy. In this paper, LBP as feature extraction is proposed with a modification of Huffman rule to enhance the feature extraction by handling the losing texture information from LBP in low resolution image. The wide dimension of feature is reduced by using the principal component analysis (PCA) method. Feature extraction is developed with aims to get the better feature of biometric representative and improve performance of palm print recognition. Extensive experiments are tested on two public databases, PolyU and CASIA database. The experimental results in low resolution image, show that this method has better palm print biometric representative for machine perspective and has the ability to improve the recognition performance with accuracy of 97.5 % (PolyU) and 92.5 % (CASIA).