Herve Lombaert
picture Algorithms

Here are some pages where I try to explain some topics related to my studies (explanations, examples, or tutorials).
Level set method: Explanation
The level set method is an elegant solution for interface tracking. This page explains what this is, and shows how to use it for segmentation.

See page here

Movies:
annealing
[76KB]
merge and split
[408KB]
circle
[336KB]
medical heart
[508KB]
custom form
[644KB]
medical knee
[1.3MB]
surface evolution
[3.2MB]
Levelset
Region based segmentation with levelsets
Edge based segmentations are intuitive since edges usually show an object boundary. However in some or many cases, edges are not sharp or are simply not there. Region based segmentations is a solution to this problem. Here is an example using a level set framework.

See page here

Movies:
leaking
[1.1MB]
region based
[880KB]
using gradient
[36KB]
closed contours
[304KB]
with noise
[168KB]
leak solved
[1.8MB]
smoothed
[152KB]
Region
Manual registration with thin plate splines
Here is a page that highlights the use of thin plate splines in image registration. Provided a few matching points, deformation can be recovered with thin plate splines.

See page here

Thinplates
Energy minimization with Graph Cuts
Beyond being a powerful segmentation technique, Graph Cuts can be used for efficient energy minimization. This page explains the rough idea of the alpha-expansion and the alpha-beta-swap algorithm. Two examples are then ilustrated. One restoring a noisy image, and one recovered the deformation map between two images.

See page here

Energy