The research of fingervein feature extraction based on the ridgelet transform

Employing multi-scale analysis to extract features in the image-based identification system has been the research focus in recent years. Among which, the relative research about adopting wavelet transform and wavelet moment as the feature have made a lot of achievements, In this paper, We employ two methods to extract finger vein feature based on the ridgelet transform because of the unsatisfactory performance of wavelet in dealing with multi-dimensional function singularity. The first kind of feature can be obtained by reducting the ridgelet coefficients' dimensionality of different scales with PCA. Although the representation of straight-line singularity using ridgelet analysis is optimal, but it's worse for curve line. So we attempt to analyze the sub-image with ridgelet, the singular angle and ridgelet coefficient statistics characteristics are chosen to construct feature vector. Finally, we employ nearest neighbor classifier to implement classification and recognition. The result show that both of the methods has their own strengths. But the second feature has a higher recognition rate with high-quality images.