Fingerprint segmentation using improved automatic labeling based linear neighborhood propagation
Due to applications of various sensors, fingerprint segmentation encounters sensor interoperability problem. The Automatic Labeling based Linear Neighborhood Propagation (ALLNP) segmentation method, which learns a segmentation model only based on the input image, is a sensor interoperable method. However, the traditional ALLNP method labels constant number of blocks barely based on contrast feature, which may inject some noise and degrade the segmentation performance. To effectively address the issue, we present a method called IALNP which makes improvement to the ALLNP. IALNP provides a more robust automatic labeling mechanic, which combines the variance with gradient magnitude to exactly label partial blocks for LNP learning. Experimental results show that our proposed method achieves higher accuracy than traditional method and simultaneously has strong adaptability to deal with sensor interoperability.