Stochastic Multiscale Segmentation Constrained by Image Content
Abstract
We introduce a noise-tolerant segmentation algorithm efficient on 3D multiscale granular materials. The approach uses a graph-based version of the stochastic watershed and relies on the morphological granulometry of the image to achieve a content-driven unsupervised segmentation. We present results on both a virtual material and a real X-ray microtomographic image of solid propellant.
Origin : Files produced by the author(s)
Loading...