Optimal Transport Mixing of Gaussian Texture Models

Sira Ferradans (1) , Gui-Song Xia (1) , Gabriel Peyré (1) , Jean-François Aujol (2)

1 CEREMADE - CEntre de REcherches en MAthématiques de la DEcision
2 IMB - Institut de Mathématiques de Bordeaux

This paper tackles the problem of mixing color texture models learned from an input dataset. We focus on stationary Gaussian texture models, also known as spot noises. We derive the barycenter and geodesic path between models according to optimal transport. This allows the user to navigate inside the set of texture models, and perform texture synthesis from the obtained interpolated models. Numerical examples on a library of exemplars show the ability of our method to generate arbitrary interpolations among unstructured natural textures.

Mots clés: Optimal transport; gaussian model; texture synthesis; texture mixing

Domaines: Traitement du signal et de l'image [eess.SP]; Traitement du signal et de l'image [eess.SP]

Sira Ferradans, Gui-Song Xia, Gabriel Peyré, Jean-François Aujol. Optimal Transport Mixing of Gaussian Texture Models. 2011. ⟨hal-00662720⟩