SLIC Superpixel-Based l2,1-Norm Robust Principal Component Analysis for Hyperspectral Image Classification

. 2019 Jan 24;19(3):479. doi: 10.3390/s19030479

Algorithm 1 SLIC superpixel segmentation.

Input:
Processed HSIs image $X \in R^{N \times M}$ , Desired number of approximately superpixels K.
1:
Initialize cluster centers $C_{k} = {[l_{k}, a_{k}, b_{k}, x_{k}, y_{k}]}^{T}$ by sampling pixels at regular grid steps S.
2:
Move cluster centers to the lowest gradient position in a $3 \times 3$ neighborhood.
3:
Set label $l (i) = - 1$ for each pixel i.
4:
Set distance $d (i) = \infty$ for each pixel i.
5:
Repeat
6:
for each cluster center $C_{k}$ do
7:
for each pixel i in a $2 S \times 2 S$ region around $C_{k}$ do
8:
Compute the distance D between $C_{k}$ and i.
9:
if $D < d (i)$ then
10:
set $d (i) = D$
11:
set $l (i) = k$
12:
end if
13:
end for
14:
end for
15:
Compute new cluster centers.
16:
Compute residual error E.
17:
Until $E \leq$ threshold.
18:
Enforce connectivity.
Output:
Superpixel segmentations.