
190 Variational Methods in Image Processing
The segmentation method introduced in [208] is base d on the geode sic
active contour model and includes a topological constraint based on a geo-
metrical observation. In the following, we a ssume that Φ is a s igned-distance
function to C
Φ(x) =
dist(x, C) if x ∈ Ω − ¯ω
−dist(x, C) if x ∈ ω .
(9.20)
In that case, Φ satisfies |∇Φ| = 1. The gradient ∇Φ is perpendicular to the
isocontours of Φ. It points in the direction of increasing Φ, s o the unit outward
normal vector to the zero level line at point x is merely ∇Φ(x) (more generally,
the unit outward normal vector to the k level line at point x is ∇Φ(x)).