Zig-Zag Based Single-Pass Connected Components Analysis

View full-text article in PMC

. 2019 Apr 6;5(4):45. doi: 10.3390/jimaging5040045

Algorithm 4 UpdateDataStructures

1:
if $I [X]$ then ▹Object pixel
2:
if $\neg L_{B}$ then
3:
if $L_{A o r D} \land L_{C} \land L_{A o r D} \neq L_{C}$ then ▹ Merger operation
4:
if $L_{A o r D} . r w \leq L_{C} . r w$ then ▹ Propagating merger
5:
$L_{X} : = L_{A o r D}$ ▹ Assign representative label
6:
$L_{o l d} : = L_{C}$
7:
$L_{C} : = L_{X}$ ▹ Update neighbourhood label
8:
else
9:
$L_{X} : = L_{C}$ ▹ Assign representative label
10:
$L_{o l d} : = L_{A o r D}$
11:
end if
12:
$M T [L_{o l d}] : = L_{X}$ ▹ Record merger in table
13:
$D T [L_{X}] : = D T [L_{X}] \circ D T [L_{o l d}] \circ I F V (X)$ ▹ Merge data (and active tags)
14:
$L_{o l d} \to L a b e l F I F O$ ▹ Recycle the old label
15:
else if $\neg L_{A o r D} \land \neg L_{C}$ then ▹ New label operation
16:
$L_{X} : = n e w L a b e l$ ( $\leftarrow L a b e l F I F O$ ) ▹ From a recycle queue
17:
$L_{X} . r w : = y$ ▹ Augment label with row number
18:
$M T [L_{X}] : = L_{X}$ ▹ Initialise merger table
19:
$D T [L_{X}] : = I F V (X)$ ▹ Start feature vector
20:
else
21:
if $L_{A o r D}$ then ▹ Copy $L_{A o r D}$
22:
$L_{X} : = L_{A o r D}$
23:
else ▹ Copy $L_{C}$
24:
$L_{X} : = L_{C}$
25:
end if
26:
$D T [L_{X}] : = D T [L_{X}] \circ I F V (X)$ ▹ Add current pixel to data table
27:
end if
28:
else ▹ Copy $L_{B}$
29:
$L_{X} : = L_{B}$
30:
$D T [L_{X}] : = D T [L_{X}] \circ I F V (X)$ ▹ Add current pixel to data table
31:
end if
32:
else
33:
$L_{X} : = 0$ ▹ Background pixel
34:
if $D T [L_{A}] . A T = X$ then ▹ Check completed object
35:
Output: $D T [L_{A}]$
36:
$L_{A} \to L a b e l F I F O$ ▹ Recycle the label
37:
end if
38:
end if
39:
$R B [X] : = L_{X}$ ▹ Save label in row buffer for next row