Abstract |
Stochastic watershed is a novel segmentation method based on a probability density function. This probability density function is created by repeated realizations of seeded watershed with a fixed number of random seed points. We study the relationship between the algorithm's parameters and attributes of the image in order to find the strengths and the weaknesses of the algorithm. The stochastic watershed works best when the regions are of similar size.
Additionally, the number of iterations can be minimized by choosing a number of markers close to the number of expected regions in the image. In the case where the regions are different sized, the original algorithm yields unsatisfactory results. |