Texture Classification By Statistical Learning From Morphological Image Processing: Application To Metallic Surfaces

Aurélien Cord (1) , Francis Bach (2, 3) , Dominique Jeulin (4)

1 IFSTTAR/COSYS/LIVIC - Laboratoire sur les Interactions Véhicules-Infrastructure-Conducteurs
2 LIENS - Laboratoire d'informatique de l'école normale supérieure
3 SIERRA - Statistical Machine Learning and Parsimony
4 CMM - Centre de Morphologie Mathématique

A classification method based on textural information for metallic surfaces displaying complex random patterns is proposed. Because these kinds of textures show fluctuations at a small scale and some uniformity at a larger scale, a probabilistic approach is followed, considering textural variations as realizations of random functions. Taking into account information of pixel neighbourhoods, the texture for each pixel is described at different scales. By means of statistical learning, the most relevant textural descriptors are selected for each application. The performance of this approach is established on a real data set of steel surfaces.

Mots clés: Mathematical morphology; rough surfaces; statistical learning; texture classification

Domaines: Matériaux

Aurélien Cord, Francis Bach, Dominique Jeulin. Texture Classification By Statistical Learning From Morphological Image Processing: Application To Metallic Surfaces. Journal of Microscopy, 2010, 239 (2), pp.159-166. ⟨10.1111/j.1365-2818.2010.03365.x⟩. ⟨hal-00836006⟩