Wavelet Decomposition Feature Parallel Fusion by Quaternion Euclidean Product Distance Matching Score for Finger Texture Verification - Advances in Computational Science and Engineering - page 147

Table 2. Computation time cost

Procedure Preprocessing Feature Extraction Matching

Computation Cost (ms) 1.91 0.03 58.43

the table, the procedure of preprocessing mainly refers to the window ROI ex-
traction described in the subsection 3.1, feature extraction includes the phase of
2-D wavelets decomposition feature extraction and construction of quaternion,
and matching time means matching between template and tester with QEPD. It
is true that computation time cost is probably reduced provided that we optimize
the code.

4.5 A Comparison with Palmprint QEPD

In the experimental section, we also propose palmprint QEPD for a comparison
of performance with such finger texture QEPD. There is a distribution of QEPD
matching score between genuine and imposter in fig.8.
From the graph, the distance between the peak of inner-class and outter-class
is shorter than that of finger texture of QEPD. If two peaks are ”closer”, it will
be more dicult to discriminant genuine and impostor. Here the peak of intra-
class is at 0.1, and inter-class 0.24. They are unlikely to separate widely like

The distribution of inner−class and outer−class QEPD

0.12

inner−class QEPD
outer−class QEPD

0.1

0.08

0.06

0.04

0.02

0

0 0.2 0.4 0.6 0.8 1

Fig. 8. The distribution of genuine and impostor palmprint QEPD matching score