Algorithm 2. The proposed SuFMoFPA approach
Input: Input image which is to be segmented
Output: Segmented output image
1: Find the gradient image corresponding to the input image using the method proposed in (Hore et al., 2015).
2: Apply Eqs. (9) and (10) to find the superpixel image corresponding to the input image.
3: Determine the representative point τ of a superpixel.
4: Randomly initialize the cluster centers $C_i={\tau }_{\mathit{low}}+random\left(0,1\right)\left({\tau }_{\mathit{high}}-{\tau }_{\text{lo}w}\right)$ where ${\tau }_{\mathit{high}}$ and ${\tau }_{\text{lo}w}$ denotes the upper and lower bound respectively for a representative point.
5: Randomly assign the fuzzy membership values to the superpixels.
6: $\mathit{nItr}\leftarrow 1$//Iteration counter
7: Repeat until $\mathit{nItr}>evalCnt$//evalCnt is the maximum number of iterations
8:  Determine the fitness values
9:  Perform global pollination
10:  Perform local pollination
11:  Update the solutions using Eq. (20)
12:  Check if $S_p^{\mathit{itr}+1}$ is worse than $S_p^{\mathit{itr}}$ then
13:   $S_p^{\mathit{itr}+1}=S_p^{\mathit{itr}}$
end if
14:  Update the global best
end until
15: Prepare the output segmented image by assigning the superpixels to their nearest cluster centers.
16: Return the segmented image.