Hierarchical Fusion Using Subsets of Multi-Features for Historical Arabic Manuscript Dating

. 2022 Mar 1;8(3):60. doi: 10.3390/jimaging8030060

Algorithm 1 Feature extraction algorithm based on hierarchical fusion approach.

Input: Raw features (views)

X^{F E_{l_{0 j}}}

j = 1, 2, \dots, n

, regularization parameter

λ

i = 0

.
Output: Fused features

X^{F E_{l_{i j}}}

1:
Compute the set of all subsets of $X^{F E_{l_{i j}}} = X^{F E_{l_{0 j}}}$ except to ∅ and with members less than two members: $P_{l_{i}} (X^{F E_{l_{i j}}})$ .
2:
repeat
3:
Select one of the subset $P_{l_{i}} (X^{F E_{l_{i j}}})$ : $S_{l_{i}}$ .
4:
Compute dictionaries set of $S_{l_{i}}$ using [31].
5:
$i = i + 1$ .
6:
Apply fusion method using (1): $X^{F E_{l_{i j}}}$ .
7:
Compute updated set of $P_{l_{i}} (X^{F E_{l_{(i) j}}})$ using (2).
8:
until ( $P_{l_{i}} (X^{F E_{l_{(i - 1) j}}}) - S_{l_{i - 1}} \neq \emptyset$ )