Algorithm 1 Observation Candidate Generation in a Region

Input:

Ωm: the image region

Sm: the set of positions on the skeleton of the region

Pm: the set of positions of the local maxima in the region {X(k1), X(k2), …, X(kN)}: the set of feasible targets

Start:

n = l

Emax = 0

For i = 2 to N

k = 0

Step 1: try to get a new subset {X(kh1), X(kh2), …, X(khi)

If a new subset is available, then go to Step 2

Else, go to Step 5

Step 2: check the subset

If the subset is feasible, then go to Step 3

Else, go to Step 1

Step 3: n = n + 1, find the observation-set ${\mathbf{D}}_{(t,{\mathbf{\Omega }}_m)}^{(m,n)}$, which maximizes Eq.(19)

(1) Search on Sm and Pm to get initial result

(2) Alow the observations to vary within 1 pixel, get refined result and the energy $E_{I_t}^{(m,n)}$

Step 4: check the observation-set ${\mathbf{D}}_{(t,{\mathbf{\Omega }}_m)}^{(m,n)}$

If $E_{I_t}^{(m,n)}>E_{\mathit{max}}$, then keep it and k = k + 1

Else, discard the observation-set and n = n − 1

Go to Step 1

Step 5: If k > 0, then $E_{\mathit{max}}=\mathit{max}\{E_{I_t}^{(m,2)},\dots ,E_{I_t}^{(m,n)}\}$ Else, break the For loop

End

Output:

observation-sets and associated subsets of the targets