Vein Pattern Recognitions by Moment Invariants

Abstract

In this paper, dyadic wavelet transform is adopted to extract finger-vein pattern from finger images, which are not only contain vein pattern but also shading and noise. Images are transformed from spatial domain to wavelet domain, and wavelet coefficients of the vein patterns and the noise are processed by soft-thresholding denoising method, which can recover the vein pattern from noisy data. Then compute modified moment invariants of the reconstruction images as the vein pattern feature to represent the vein pattern features. Vein pattern features matching bases on Hausdorff distance. Experiment results show that this method is stabile and fast for extracting vein pattern from noisy data.