We have presented a variational model based on level set formulation
for image classification. The level set formulation is a way to
represent regions and set of interfaces with a continuous function defined
over the whole support of the image. The minimization of the functional
leads to a set of coupled PDE's which are considered through a dynamical
scheme. Each PDE is guiding a level set function according to
internal forces (length minimization), and external ones (data term,
no vacuum and no region
overlapping). Results on both synthetic and satellite images are
- Further work will be conducted to deal with the estimation
of the class parameters (unsupervised classification). We also
envisage to extend this model to multispectral data (with applications
to multiband satellite data and applications to color imaging).
- More details can be found in the research report.
Begining of the demo
Back to the