# Voxel filler

This page describes an program for reconstructing neurons from membrane outlines. It uses a fast marching algorithm, a 3D extension of the one in: 2D fast marching.. The test data comes from a 3D neuron tissue grower.

The algorithm works by:

- Calculating the gradient at each point in voxel space.
- Picking a number of seed points as "alive"
- "tasting" the neightbouring points to find the path of least resistance
- Growing in that direction, and repeating

It gains over a simple "flood fill" as it expands at the *global
minimum* of the boundary of the already filled area - making
it less likely to breach cell boundaries if there is more cell to fill.

Some of the problems for the algorithm:

- "Gaps" in the membrane - the fill can leak through.