High-resolution Palmprint Minutiae Extraction Based on Gabor Feature
JuFu Feng,ChongJin Liu,Han Wang,Bing Sun
DOI: https://doi.org/10.1007/s11432-014-5125-5
2014-01-01
Science China Information Sciences
Abstract:Extracting effective minutiae is difficult for high-resolution palmprint, because of the strong influence from principal lines, creases, and other noises. In this paper, a novel minutiae detection and reliability measurement method is proposed for high-resolution palmprint minutiae extraction. Firstly, we propose the Gabor Amplitude-Phase model for palmprint representation, which contains sufficient palmprint information and consists of the phase field and amplitude field. Because of the explicit meanings of minutiae in phase field, a minutiae descriptor is constructed to detect them directly. Also, to measure minutiae reliability and remove the unreliable ones, the Gabor Amplitude-Phase feature vector is designed. It can be used for describing the local area of a minutia redundantly. Then, the Adaboost algorithm is introduced in model training to select best features and corresponding weak classifiers for minutiae authenticity discriminant. Finally, the response value of weighted linear combination of selected weak classifiers is used for minutiae reliability measurement and unreliable ones removal. According to our analysis, the selected features are meaningful and useful for describing the minutiae area and measuring their reliability. Experimental results show that our proposed method is effective for minutiae extraction and can improve the matching performance.