Data clustering by optimum-path transform / Agrupamento de dados por florestas de caminhos otimos




This thesis proposes a novel approach for data clustering based on optimum-path forest. The samples are taken as nodes of a graph, whose arcs are defined by an adjacency relation. The nodes are weighted by their probability density values (pdf) and a connectivity function is maximized, such that each maximum of the pdf becomes root of an optimum-path tree (cluster), composed by samples "more strongly connected" to that maximum than to any other root. Irrelevant maxima (domes) are removed by morphological reconstruction of the pdf. A discussion shows the advantages over aproaches based on the gradient of the pdf and its relations with the watershed transform. The method is extended for large datasets and compared to mean-shift and watershed approaches for pixel clustering and interactive segmentation, respectively. The experiments for the segmentation of natural images and automatic gray and white matter classification of the brain show fast and accurate results compared to state-of-the-art methods.


image analysis sistemas de recuperação de informação processamento de imagens information retrieval systems banco de dados image processing databases analise de imagem

Documentos Relacionados