``Anisotropic_Diffusion'' -- anisotropic diffusion filter

purpose:

Anisotropic diffusion filtering is mathematically formulated as a diffusion process. The anisotropic filter smoothens regions within boundaries in preference to smoothing across the boundaries, thus ensuring that the boundaries of regions in the image remain sharp.

input:

An 8-bit volume.

output:

An 8-bit volume.

parameters:

Default parameters menu is

Dimension = 2D
Direction (For 2D only) = TRANSVERSE
Gradient type = Exponential
Number of iterations = 1
Scaling parameter, K = 1
lambda(max_2D = 0.25; max_3D = 0.037) = 0.25
input from vol # = 0
copy output to vol # = 0

comments:

 

1. This is a moderately fast operation using the 2D filter. 3D filtering could be slow for large number of iterations.
2. It is difficult to estimate the number of iterations required for a desired level of smoothing. This is a function of the gradient type, scaling parameter K, and the diffusion rate factor. For the 3D filter, a smaller value of K and lesser number of iterations are typically needed to achieve (approx.) the same level of blurring as in the 2D case. This is due to the fact that 26 neighbors are used in the 3D filter as compared to 4 neighbors in the 2D filter.