Graphics Reference
In-Depth Information
Fig. 9.15. Closed snake being prematurely stopped by a local minimum when
applied to cell image segmentation problem (from [11]).
Each snake in the network is evolved within its search space as an open-
ended snake, without regard to its connectivity to other snakes in the network
to estimate the optimal position of its mid-points. Next, each snake is split
in two at the mid-points to create two new half-snakes. We fix the location
of the mid-point end of each half-snake and perform a forward pass of the
Viterbi algorithm. This yields the minimum energy of each half-snake for all
m possible positions of its joint-end. All half-snakes that are connected by a
joint have the same m possible positions for their joint-end so we can then
determine the common joint-end position that minimizes the total energy of
Search WWH ::




Custom Search