Next: Complexity and speeding up
Up: Segmentation
Previous: Power series expansion
Contents
We can then define a segmentation algorithm, based on the restoration algorithm previously defined in section 2.4:
-
- Solve the direct (2.21) and adjoint (2.27) unperturbed problems with
everywhere.
-
- Compute the
matrix
defined by equation (2.28) and its lowest eigenvalue
at each point of the domain
.
-
- Define
the edge set, where
is a small negative threshold.
-
- Set
the minimal value of
for which it is easy to compute numerically the solution
of problem
.
-
- Choose
in order to have an approximation error in
, and choose
different values
.
-
- Compute the solutions
in
of problems
.
-
- Compute the interpolation polynomial
of degree
, defined by equation (2.52), for
.
This algorithm has a complexity in
, where
is the number of pixels in the image, and
is the degree of the interpolation approximation. In numerical experiments,
is typically of the order of
to
.
Several numerical tests are detailed in [18].
Next: Complexity and speeding up
Up: Segmentation
Previous: Power series expansion
Contents
Back to home page