Unsupervised multi-level segmentation of multispectral images

Abstract

The authors describe a scheme that performs multilevel segmentation of an image at many scales using a multiresolution texture representation. Each level uses anisotropic diffusion to segment a multispectral image at successively lower resolutions. Texture and statistical similarities between and within levels guides the diffusion process. The restriction of coarse-to-fine segmentation is removed, and one operates at all levels simultaneously. In this manner the labeling process can choose the scale or scales at which useful segments exist. The network outperforms a fuzzy clustering scheme in the segmentation of a high-resolution multispectral aerial image.<>