Algorithm 1 Sequence matching using feature CSLBP++GIST.

Inputs:

${\{I_j^{train}\}}_{j=1}^{N^{train}}$ {training images}; ${\{I_i^{test}\}}_{i=1}^{N^{test}}$ {testing images };

$N^{train},N^{test}$ {training and testing images numbers}; $L_{length}$ {Sequence length};

Outputs:

D{distance matrix};

Algorithm:

$/\ast$ TRAINING PHASE $\ast /$

for j← 1 to $N^{train}$/$L_{length}$ do

$F_j^{train}$←$f_{CSLBP}++f_{GIST};$$//$Feature of training sequence.

end for

$/\ast$ TESTING PHASE $\ast /$

/∗ Sequence matching based on feature sequences; Section 3.2 ∗/

for i← 1 to $N^{test}/L_{length}$ do

$F_i^{test}$←$f_{CSLBP}++f_{GIST};$$//$Feature of testing sequence.

for j← 1 to $N^{train}/L_{length}$ do

$D_{i,j}\leftarrow {\sum }_k\frac{{({\left(F_i^{test}\right)}_k-{(F_j^{train})}_k)}^2}{|{\left(F_i^{test}\right)}_k+{\left(F_j^{train}\right)}_k|}$; $//$k is the index of the components of feature vector (See Equation (7)).

end for

end for