Distance Transform

Compute the generalized distance transform of a sampled function

Introduction

This module provides a Python implementation of the linear-time distance transform described in:

P. Felzenszwalb, D. Huttenlocher "Distance Transforms of Sampled Functions"

Computing the distance transform is as easy as:

import dt
import numpy as np

x = np.random.standard_normal((100,100))
y,i = dt.compute(x)

This module can handle arbitrary dimensional data:

x = np.random.standard_normal((100,100,4,5))
y,i = dt.compute(x)  # compute the distance transform across ALL dimensions
y,i = dt.compute(x, axes=(0,1)) # Compute across the (0,1) axes in the tensor

You can also change the distance function, or parameters used:

y,i = dt.compute(x, f=dt.L2(0.01)) # reduce the distance penalty

Installing

Install the package using pip:

pip install git+https://github.com/hbristow/distance-transform

You will need Cython to build the extensions.