Fast Bilateral Filter for 3D Images

Fast bilateral filter for 3D images

This is an implementation of fast bilateral filtering for 3D images. This filter smoothes the image while preserving edges, but in its most straightforward implementation is very computationally demanding, especially with large 3D images. Fast bilateral filtering (S. Paris and F. Durand, MIT[1]) is an approximation technique which drastically improves the speed of computation. My implementation closely follows the Matlab version of the fast 2D filter[2] with some modifications.

The core function allows to control all of the parameters of the filter (4 for the Gaussian kernel, 4 for the subsampling), and a wrapper function makes some simplifications for ease of use.

To perform the separable 4D convolution, I used my own function submitted separately to File Exchange (27957).

[1] http://people.csail.mit.edu/sparis/bf/
[2] http://people.csail.mit.edu/jiawen/software/bilateralFilter.m