Table 4.

Algorithm A1

Globally adjust the intensity of CBCT

Initialize the moving vector r to zero

Down-sample the images to the coarsest resolution

Repeat for each resolution level

while l(k − 10) − l(k) ≤ ε, do

1. Compute a and b for each voxel using patches in ${\mathit{I}}_m^{(k)}$ and Is (Eqs. (5), (6));

2. Interpolate/extrapolate a and b to where ξ = 1;

3. Obtain I′s by applying the estimated linear transformation at each voxel;

4. Compute dr for each voxel (Eq. (1));

5. Add dr to the total moving vector r;

6. Regularize r by applying a Gaussian kernel;

Up-sample the moving vector r to a finer resolution level

Until the finest resolution is reached