Finger-vein image matching based on adaptive curve transformation.

Extracting reliable finger-vein features directly from original finger-vein images is not an easy task since the captured finger-vein images are always poor in quality. This paper proposes an effective method of finger-vein feature representation based on adaptive vector field estimation. Considering that the vein networks consist of vein curve segments, a set of spatial curve filters (SCFs) with variations in curvature and orientation are first designed. To fit vein curves locally and closely, SCFs is then weighted using a variable Gaussian model. Due to the fact that finger veins vary in diameters naturally, an effective curve length field (CLF) estimation method is proposed to make weighted SCFs adaptive to vein-width variations. Finally, with CLF constrain, vein vector fields(VVF) are built for finger-vein network feature description. Experimental results show that the proposed method is highly powerful in improving finger-vein matching accuracy. HighlightsA set of spatial curve filters (SCFs) are designed using a variable curve model in curvature and orientation.A Gaussian weighted curve model is proposed to reduce filtering errors in fitting vein diameters.An efficient method is proposed for reliably estimating curve length fields (CLF). This can make SCFs adaptive to vein-width variations.