4. O. Duchenne, J-Y. Audibert: Fast interactive segmentation using color and textural information
This work presents an efficient method for image multi-zone
segmentation under
human supervision. We combine different techniques to produce an intelligent
segmentation. As most of the segmentation methods, our algorithm is based on
an energy minimization. This energy contains a data-dependent term and a cohesion
term.
Our data-consistent term is based on a SVM method designed in order to deal
with color and textural information.
For the coherence term, we use a classic Canny edge detection algorithm to
provide another good but not perfect source of information about the limits
of the
segments. Finally, to minimize the energy, we use a Graph Cut algorithm.
We compare our results on an image data bank available on-line.