Algorithm for ORB-TPLGD

Input: Image data

Output: Feature description vector for each feature point

1: Extract feature points set $P=\left\{p_1,p_2,\dots ,p_i,\dots ,p_N\right\}$ from input image

2: for $i=1,2,\dots ,N$ do

3:   Define neighborhood $W$ of $M\times M$ around feature point $p_i$ 4:    In W, T three-patch groups are selected using Gaussian function, coordinates of the center point is $Z={\left\{z_t\right\}}_{t=1\dots T}={\left\{\left[\left(x_{t1},y_{t1}\right),\left(x_{t2},y_{t2}\right),\left(x_{t3},y_{t3}\right)\right]\right\}}_{t=1\dots T}$

5:   Define $k\times k$ size patches $S_{t1},S_{t2},S_{t3}$ centered on point $\left(x_{t1},y_{t1}\right),\left(x_{t2},y_{t2}\right),\left(x_{t3},y_{t3}\right)$, $S={\left\{S_t\right\}}_{t=1\dots T}={\left\{\left[S_{t1},S_{t2},S_{t3}\right]\right\}}_{t=1\dots T}$

6:   for $j=1,2,\dots ,T$ do 7:      if $f\left(S_{t1}\right)<f\left(S_{t2}\right)\&f\left(S_{t1}\right)<f\left(S_{t3}\right)$ then bit = 1 8:      else then bit = 0

9:      end if

10:   end for (obtain a binary string $b_W$) 11:   Calculate the difference within all three pixel patches groups $Q={\left\{Q_t\right\}}_{t=1\dots T}={\left\{Q_{t1}=\left|f\left(S_{t1}\right)-f\left(S_{t2}\right)\right|,Q_{t2}=\left|f\left(S_{t1}\right)-f\left(S_{t3}\right)\right|\right\}}_{t=1\dots T}$       Calculate the mean of all differences $Q_{average}$

12:    for $j=1,2,\dots ,T$ do

13:      if $Q_{average}<Q_{t1}\&Q_{average}<Q_{t2}$ then bit = 1 14:      else then bit = 0 15:      end if

16:   end for (obtain a binary string $b_w^{^}$) 17:     $b_W$ and $b_w^{^}$ are fused to get final $Descriptio{n}_{keypoint}$

18: end for