Next: Neural
Net Design Issues Up: Main
page Previous: An
MRI Application
Neural Networks for Boundary
Tracing
A model of our interaction scenario is illustrated on an enlarged set of
pixels, shown in Figure 3. The black pixels represent a trace of 120 pixels.
The red pixels on the left were traced manually with a cursor over the
image; that pixel segment became the neural nets exemplar to learn its
contour, and then project the trace along the contour learned.
Figure 3: Enlargement: Network-traced path through the grey-scale
landscape.
The contour is learned and followed by tracking characteristics of pixels
to the left, to the right, and on the directed path, as in Figure 4.
Figure 4: A path and its neighbors
The neural net evaluates possible next pixels on the contour, based
on what it has been trained on in the past.
Next: Neural Net Design
Issues Up: Main page
Previous: An MRI Application
stewart crawford-hines