Fingerprint Smear Detection Based on Subband Feature Representation

Abstract

Fingerprint smear detection has become a challenging issue due to the erratic texture of the smear tissue and its similarity to normal finger area. This paper presents a novel fingerprint image smear detection approach integrating symmetric wavelet transform (SWT), gray level co-occurrence matrix, and DCT. A feature extraction algorithm is first proposed by utilizing SWT to decompose each fingerprint and characterizing local texture features of defective finger tissue with the SWT coefficients in subbands 4~19. Concurrence matrix-based texture features are incorporated into the feature vector to further improve the texture classification sensitivity. The concatenated feature vector is then fed into a pretrained genetic neural network classifier, which identifies smears by labeling fingerprint subblocks into different categories. Finally, DCT decomposition is used to detect abnormalities in fingerprint images containing small smear areas and abrupt breakages. Experimental results indicate that the hybrid method can effectively identify various types of fingerprint smears.

Publisher note

To access the full article, please see PDF.