Skip to Main content Skip to Navigation
Theses

Modélisation d'objets 3D par fusion silhouettes-stéréo à partir de séquences d'images en rotation non calibrées

Abstract : We present a new approach to high quality 3D object reconstruction using stereo and silhouettes. Starting from an uncalibrated sequence of color images under circular motion, the algorithm is able to calibrate the camera (both the pose and the focal length), and reconstruct the 3D geometry and the texture. As a difference with other classic methods, calibration is accomplished using a set of silhouettes as the single input data. We develop the concept of silhouette coherence applied to a set of silhouettes generated by a 3D object. We first discuss the silhouette coherence notion and define a practical criterion to estimate it. This criterion depends both on the silhouettes and on the parameters of the cameras that generated them. These parameters can be optimized for the 3D modeling problem by maximizing the overall silhouette coherence. The core of the reconstruction method is based on a deformable model, which defines the framework where texture and silhouette information can be fused. This is achieved by defining two external forces based on the images: a texture driven force and a silhouette driven force. The texture force is computed in two steps: a multi-stereo correlation voting approach and a gradient vector flow diffusion. Due to the high resolution of the voting approach, a multi-grid version of the gradient vector flow has been developed. Concerning the silhouette force, a new formulation of the silhouette constraint is derived. It provides a robust way to integrate the silhouettes in the evolution algorithm. As a consequence, we are able to recover the contour generators of the model at the end of the iteration process. Finally, a texture map is computed from the original images for the reconstructed 3D model. The proposed 3D modeling system has been successfully validated with the reconstruction of a great variety of high quality models.
Complete list of metadatas

https://pastel.archives-ouvertes.fr/pastel-00000862
Contributor : Ecole Télécom Paristech <>
Submitted on : Monday, November 22, 2010 - 5:15:35 PM
Last modification on : Friday, July 31, 2020 - 10:44:05 AM
Long-term archiving on: : Friday, October 19, 2012 - 1:20:17 PM

Identifiers

  • HAL Id : pastel-00000862, version 1

Collections

Citation

Carlos Hernández Esteban. Modélisation d'objets 3D par fusion silhouettes-stéréo à partir de séquences d'images en rotation non calibrées. domain_other. Télécom ParisTech, 2004. Français. ⟨pastel-00000862⟩

Share

Metrics

Record views

547

Files downloads

1772