A ten-print card segmenting algorithm

Nguyễn Thị Hương Thủy, Nguyễn Ngọc Minh, Nguyễn Ngọc Kỷ
2012 Journal of Computer Science and Cybernetics  
This paper proposes a novel algorithm for segmenting fingerprints from ten-print card images. The proposed algorithm is constituted of two stages. The coarse stage uses the normalization, the smoothing, the binary transformation and then the contour following technique to cutting out a ten-print card image into twenty small fingerprints. Then, the fine stage uses the mean, the variance, the length and the directional features of the ridge segments obtained after vectorisation and piecewise
more » ... r approximation to more detail detecting the good quality fingerprint region from the noisy regions. The experimental results providing on FVC2004 DB2 and C@FRIS DB showed that the proposed algorithm worked more efficiently and gained the better precision in comparison with other methods using raster approach.
doi:10.15625/1813-9663/26/3/519 fatcat:5rbr3y2wmbdrtbmey7o2hx2rcu