A Novel Local Human Visual Perceptual Texture Description with Key Feature Selection for Texture Classification

Algorithm 1

The algorithm for PCA-based feature selection.

(1) Initialize L = 1, the number of selected features, , the selected features

set, ortho = 0, the orthogonality of the principal components computed from the selected

features, k is the principal component coefficients matrix with p rows corresponding to the

original features.

(2) Perform p - 1 times:

(i) Calculate the set s of possible combinations by choosing L features from the original p features

and the number of combinations .

(ii) Initialize i = 1, the order of the combination in the set s.

(3) Perform times:

(i) Calculate the orthogonality of the subset consisting of the i-th combination of rows in

the principal component coefficients matrix k.

(ii) If > ortho, update the orthogonality by , update the selected features

set by . Else, skip.

(iii) Update the order by .

(iv) Update the number of selected features by .

(4) Return the selected features set .