Skip to Main content Skip to Navigation

Reconstruction et analyse automatiques pour le rééclairage d'objets basés-image

Abstract : Relighting real objects from images is a difficult problem since a lot of parameters have to be estimated. Nevertheless such an operation is essential for integrating real objects into virtual scenes, for instance in the context of movies, advertisements or video games. The work proposed in this report describes a methodology for automatically reconstructing all stages of real objects estimation : object acquisition, geometry reconstruction, light sources detection, reflectance recovery and relighting. The objects we use are only represented by a series of calibrated images, they can be composed by diffuse, glossy surfaces and textures. The acquisition system, composed by a digital camera and a turntable, is robust and no additional object or pattern is used during the acquisition phase. We reconstruct a discrete geometric model and a triangular mesh having similar topological properties (closed surface, number of connected components). Multiple ponctual and directionnal light sources can be detected. Surface reflectance properties are estimated on each mesh by a parametric BRDF model. The estimated information is used for visualizing objects with new illumination conditions. New images are produced in a few seconds with a ray-tracing method.
Complete list of metadata

Cited literature [66 references]  Display  Hide  Download
Contributor : Bruno Mercier <>
Submitted on : Monday, May 7, 2007 - 10:27:38 PM
Last modification on : Wednesday, September 5, 2018 - 1:30:05 PM
Long-term archiving on: : Wednesday, April 7, 2010 - 12:42:35 AM


  • HAL Id : tel-00145124, version 1



Bruno Mercier. Reconstruction et analyse automatiques pour le rééclairage d'objets basés-image. Autre [cs.OH]. Université de Poitiers, 2006. Français. ⟨tel-00145124⟩



Record views


Files downloads