Ultimate levelings with strategy for filtering undesirable residues based on machine learning

Wonder Alves (1) , Charles F Gobber , Dennis da Silva , Alexandre Morimitsu (2) , Ronaldo F Hashimoto , Beatriz Marcotegui (3)

1 UNINOVE - Nove de Julho University [Bauru]
2 USP - Universidade de São Paulo = University of São Paulo
3 CMM - Centre de Morphologie Mathématique

Ultimate levelings are operators that extract important image contrast information from a scale-space based on levelings. During the residual extraction process, it is very common that some residues are selected from undesirable regions, but they should be filtered out. In order to avoid this problem some strategies can be used to filter residues extracted by ultimate levelings. In this paper, we introduce a novel strategy to filter undesirable residues from ultimate levelings based on a regression model that predicts the correspondence between objects of interest and residual regions. In order to evaluate our new approach, some experiments were carried out with a plant dataset and the results show the robustness of our method.

Domaines: Informatique [cs]; Traitement des images [eess.IV]

Wonder Alves, Charles F Gobber, Dennis da Silva, Alexandre Morimitsu, Ronaldo F Hashimoto, et al.. Ultimate levelings with strategy for filtering undesirable residues based on machine learning. 14th International Symposium on Mathematical Morphology and Its Applications to Signal and Image Processing, Jul 2019, Saarbrücken, Germany. ⟨10.1007/978-3-030-20867-7_23⟩. ⟨hal-02430523⟩