A Comparison between Quaternion Euclidean Product Distance and Cauchy Schwartz Inequality Distance - Advances in Computational Science and Engineering - page 132

Where k is number of pixel in such sub-images, n is the component number
of discrete signal sequence, e.g. k = 4 because of the quaternion. Notice that
D CSID ,thesumof D pixel , evidently has a characteristic of D CSID ≥

0.

3 Comparison by Experiment

3.1 Database

BJTU-HA biometric database, an inherited collection work by Institute of In-
formation Science, Beijing Jiaotong University, is utilized for our palmprint and
middle finger texture verification experiment. It contains totally 1,500 samples
from 98 person's palmprint and middle finger with different illumination condi-
tions. In the experiment, we use a subset of the database with 10 samples from
first 40 persons, totally 400 images, as our matching set.

3.2 QEPD Computation

After quaternion fusion, we employ QEPD for matching. Suppose A and B as two
quaternion vectors of palmprint or finger texture feature, A = ,

where a, b, c, d are wavelets decomposition coecients respectively. The same is
true of B =

{a + bi + cj + dk}

{t + xi + yj + zk}

. According to this distance, two matching score
tables (Table 1 and Table 2) are listed from the following palmprint (figure 1) and
finger texture ROI images (figure 2) respectively as matching examples:
Sample matching in the database is proposed for investigate the performance
of our scheme. That is, each sample is matched with other samples in the subset

Fig. 1. Palmprint ROI sample for matching. In the experiment, each ROI sample with
a 128 × 128 dimensionality has been used.