Efficient Unsupervised Image Segmentation by Optimum Cuts in Graphs

Object segmentation by Oriented Image Foresting Transform with connectivity constraints

“…The new method, successfully incorporates connectivity constraints on OIFT, preserving its low time complexity O(N = V ) (when Q is implemented using bucket sorting [9]), since it requires only four executions of the IFT algorithm. (2) The theoretical analysis of a particular case of the cost function of the Riverbed method that guarantees an optimal result according to a graph-cut measure [25]. (3) The theoretical analysis of the 2 These images are released under Creative Commons CC0 into the public domain, available at the web site https://pixabay.com/.…”

“…4The design of three new ground truth datasets from 280 public images 4 , which contain objects with thin and elongated parts, available to the community 5 . (5) Four conference papers were published in international events of high regard [19], [21], [25], [26]. This work also received the best doctoral thesis award at WVC2018 -XIV Workshop de Visão Computacional.…”

Segmentação não supervisionada de imagens via Transformada Imagem-Floresta Orientada sujeita a restrições de alto nível

“…• Um novo método de segmentação não supervisionado, chamado UOIFT (Unsupervised Oriented Image Foresting Transform), baseado em cortes ótimos em grafos dirigidos, publicado no 23rd Iberoamerican Congress on Pattern Recognition (CIARP 2018), Madrid, Espanha [Bejar et al (2018)].…”