A 2.5d approach to skin wrinkles segmentation

Etienne Decencière (1) , Amira Belhedi (1) , Serge Koudoro (1) , Frédéric Flament (2) , Ghislain François (2) , Virginie Rubert (2) , Isabelle Pécile (2) , Julien Pierre (2)

1 CMM - Centre de Morphologie Mathématique
2 L’Oreal Research and Innovation

Wrinkles or creases are common structures on surfaces. Their detection is often challenging, and can be an important step for many different applications. For instance, skin wrinkle segmentation is a crucial step for quantifying changes in skin wrinkling and assessing the beneficial effects of dermatological and cosmetic anti-ageing treatments. A 2.5D approach is proposed in this paper to segment individual wrinkles on facial skin surface described by 3D point clouds. The method, based on mathematical morphology, only needs a few physical parameters as input, namely the maximum wrinkle width, the minimum wrinkle length, and the minimum wrinkle depth. It has been applied to data acquired from eye wrinkles using a fringe projection system. An accurate evaluation was made possible thanks to manual annotations provided by three different experts. Results demonstrate the accuracy of this novel method.

Mots clés: wrinkle segmentation; mathematical morphology; 3D point cloud; skin

Domaines: Traitement des images [eess.IV]; Réseau de neurones [cs.NE]

Etienne Decencière, Amira Belhedi, Serge Koudoro, Frédéric Flament, Ghislain François, et al.. A 2.5d approach to skin wrinkles segmentation. Image Analysis & Stereology, 2019, 38 (1), pp.75. ⟨10.5566/ias.1925⟩. ⟨hal-02428880⟩