[Insight-users] GeodesicActiveContoursLeveSetImageFilter

[Koen Vermeer]

Hi Eduardo,

On Mon, 2009-01-26 at 16:02 -0200, Eduardo - LNCC wrote:

> Anybody knows why does GeodesicActiveContoursLeveSetImageFilter need a
> second image like input (SetFeatureImage() method)?

I'm not sure I really get your question. With any level set method, you
need at least two input images: The initial level set, and something for
the level set operators to work on.

For the GeodesicActiveContoursLevelSetImageFilter, the latter is often a
(smoothed) gradient image, in which the areas where the level set should
stop propagating are near 0, and the areas where the level set should
move fast are near 1. This image is then used for the location-dependent
weighing of the curvature, for the propagation speed, and indirectly for
the advection (which is calculated by taking the negative of the
derivative of the feature image).

Best,
Koen