[Insight-users] Distance field of an image

[Vikram Chalana]

Zein, Mark, 

  The FastMarchingImageFilter should work faster than Danielsson if you do
not need distances everywhere on the image. In the fast marching algorithm
you are able to specify a maximum distance beyond which you do not care to
compute. 
  
  However, if you need distances everywhere on the image, then I suspect
Danielsson filter might be faster - I would be interested in knowing if
anyone has benchmarked it though...

   Also, another couple of factors to consider (1) FastMarching gives better
approximation to true Euclidean distances compared to the Danielsson filter,
but (2) Danielsson filter in ITK provides a closest point transform for free
- it gives you the offset to the closest points, and FastMarching filter
does not. You could, however, potentially use auxiliary variables in
FastMarching to get that - but its more work.. 

Vikram