Palmprint Feature Extraction Based on Multi-wavelet and Complex Network.
Lijian Zhou,Chen Zhang,Zhe-Ming Lu,Kun Zhao,Tingyuan Nie
2017-01-01
Abstract:A new palmprint feature extraction method is proposed by analyzing the characteristics of palmprint images, multi-wavelet analysis and complex network. As we all know, the main lines of a palmprint image is a key to recognize a palmprint, and the other thin and small lines usually may influence the recognitions results relatively small. Motivated by this, the principle line feature extraction is preferred in the palmprint recognition. Multi-wavelet analysis can not only extract the coarse information, but also give the detail information in horizontal and vertical direction. By analyzing the multiwavelet features of palmprint images, it can be seen that the low frequency sub-band image with a multi-wavelet decomposition can present the principle lines very well. So the low frequency component is selected first. By observing the 4 components of the low frequency sub-image, three components except for the diagonal component contain almost all information, so these components are extracted as the elementary feature of a palmprint image. Furthermore, the binarization computation is performed on the 3 extracted features with different thresholds to extract the secondary feature of a palmprint image. Instead of directly using the binary features, the connectivity of the points in the principle line is considered because the relative correlation of the principal lines is also a key factor to recognize a palmprint image. Therefore, a series of dynamic evolution smallworld networks of the feature points in the principal line are built and the final complex network features are extracted, which contain maximum degree, average degree, standard deviation of the degree and contrast. Experimental results on the CASIA Palmprint Image Database show that the proposed method is effective, especially for the little training sample number.