A New Feature Set for Texture-Based Classification of Remotely Sensed Images in a Quantum Framework

. 2026 Mar 30;12(4):149. doi: 10.3390/jimaging12040149

Algorithm 1 SVGM Texture Feature Extraction

1:
Input: Set of labeled images ${I_{1}, I_{2}, \dots, I_{N}}$ , patch size p, threshold levels t
2:
Output: Feature matrix F for classification
3:
for each image $I_{i}$ in dataset do
4:
Divide $I_{i}$ into non-overlapping patches of size $p \times p$
5:
Store all patches in set $P$
6:
end for
7:
Shuffle $P$ randomly
8:
Construct collage image C by arranging patches from $P$ in grid layout
9:
Apply Otsu thresholding to C to obtain $C_{t}$ with t levels
10:
Initialize feature matrix $F \leftarrow \emptyset$
11:
for each patch $P_{j}$ in $C_{t}$ do
12:
Compute GLCM matrices $G_{0}$ , $G_{45}$ , $G_{90}$ , $G_{135}$ for directions $0 °$ , $45 °$ , $90 °$ , $135 °$ and $d = 1$
13:
Aggregate GLCMs: $G \leftarrow \frac{1}{4} (G_{0} + G_{45} + G_{90} + G_{135})$
14:
Perform SVD: $S V D (G) = U Σ V^{T}$
15:
Extract top-t singular values from $Σ$ as feature vector $f_{j}$
16:
Append $f_{j}$ to feature matrix F
17:
end for
18:
Return F