A rigorous and realistic Shape From Shading method and some of its applications
نویسندگان
چکیده
This article proposes a rigorous and realistic solution of the Lambertian Shape From Shading (SFS) problem. The power of our approach is threefolds. First, our work is based on a rigorous mathematical method: we define a new notion of weak solutions (in the viscosity sense) which does not necessarily requires boundary data (contrary to the work of [28, 27, 26, 4, 11]) and which allows to define a solution as soon as the image is (Lipschitz) continuous (contrary to the work of [22, 9]). We prove the existence and uniqueness of this (new) solution and we approximate it by using a provably convergent algorithm. Second, it improves the applicability of the SFS to real images: we complete the realistic work of [26, 30], by modeling the problem with a pinhole camera and with a single point light source located at the optical center. This new modelization appears very relevant for applications. Moreover, our algorithm can deal with images containing discontinuities and black shadows. It is very robust to pixel noise and to errors on parameters. It is also generic: i.e. we propose a unique algorithm which can compute numerical solutions of the various perspective and orthographic SFS models. Finally, our algorithm seems to be the most efficient iterative algorithm of the SFS literature. Third, we propose three applications (in three different areas) based on our SFS method. Key-words: Applications of Shape from Shading; real images; Shape from Shading without boundary data, viscosity solutions with state constraints; black shadows; light source located at the optical center; perspective Shape From Shading; “generic” Hamiltonian. http://www-sop.inria.fr/odyssee/team/Emmanuel.Prados/ (Odyssée Lab., INRIA/ENS/ENPC, France) http://www-sop.inria.fr/odyssee/team/Olivier.Faugeras/ (Odyssée Lab., INRIA/ENS/ENPC, France) “Shape-From-Shading”: Une approche réaliste et rigoureuse; exemples d’applications. Résumé : Cet article propose une solution réaliste et rigoureuse du problème du “Shape-FromShading” Lambertien. La puissance de notre approche est triple. Tout d’abord, celle-ci est basée sur une méthode mathématique rigoureuse: nous définissons une nouvelle notion de solutions faibles (au sens des solutions de viscosité) qui ne requiert pas nécessairement de données au bord (contrairement aux travaux de [28, 27, 26, 4, 11]) et qui permet de définir une solution dès que l’image est Lipschitzienne (contrairement aux travaux de [22, 9]). Nous prouvons l’existence et l’unicité de cette (nouvelle) solution et nous l’approximons en utilisant un algorithme dont nous prouvons la convergence. Deuxièmement, notre approche augmente l’applicabilité du “Shape-From-Shading”. Nous complétons la modélisation réaliste proposée par [26, 30], en considérant une caméra de type sténopé et une source de lumière ponctuelle disposée au centre optique: cette nouvelle modélisation est très pertinente pour les applications et permet d’obtenir d’excellents résultats sur des images réelles. Notre algorithme permet de plus de traiter des images contenant des discontinuités et des ombres portées. Celui-ci est très robuste aux bruits pixéliques et aux erreurs faites sur les paramètres. Il est aussi générique: i.e. nous proposons un unique algorithme qui peut calculer des solutions numériques des divers modèles de “Shape-From-Shading” orthographiques et perspectifs. Enfin, notre algorithme semble être l’algorithme itératif le plus “performant” de la littérature du “Shape-FromShading”. Troisièmement, nous proposons trois applications (dans trois domaines différents) basées sur notre méthode. Mots-clés : Applications du Shape from Shading; images réelles; Shape from Shading “sans” condition aux limites, solutions de viscosité avec contraintes d’état; ombres portées; source de lumière au centre optique; Shape From Shading perspectif; Hamiltonien “générique”. A rigorous and realistic Shape From Shading method 3
منابع مشابه
Shape From Shading
Shape From Shading is the process of computing the threedimensional shape of a surface from one image of that surface. Contrary to most of the other three-dimensional reconstruction problems (for example, stereo and photometric stereo), in the Shape From Shading problem, data are minimal (we use a single image!). As a consequence, this inverse problem is intrinsically a difficult one. In this c...
متن کاملUsing hybrid approaches to solve the challenges of shape from shading
For over thirty years researchers have been trying to solve the shape from shading problem of determining 3D shape from a single image with a single light source. The basic problem of determining shape from shading is made more difficult due to challenges of light orientation, camera type, ambiguity, multiple materials, and specular highlights. This paper shows how some of these challenges can ...
متن کاملFast Marching Method for Shape from Shading under Perspective Projection
The fast marching method has recently been applied to the shape from shading problem for orthographic projection [4]. Unlike many previous researches, the method is stable and monotone, and offer a well posed solution to the shape from shading problem. In this paper, the method is extended to perspective projection using a model that accounts for the foreshortening effect of perspective project...
متن کاملGarment Modeling from a Single Image
Modeling of realistic garments is essential for online shopping and many other applications including virtual characters. Most of existing methods either require a multi-camera capture setup or a restricted mannequin pose. We address the garment modeling problem according to a single input image. We design an all-pose garment outline interpretation, and a shading-based detail modeling algorithm...
متن کاملUser-Defined Shading Models for VR Applications
Quite a few techniques have been proposed on how to implement more complex and realistic shading models with graphics hardware [5, 7], making them useful for VR applications. Still, these techniques are rarely used, probably due to two reasons: complex implementation issues, and unintuitive parameters for the used shading models. We propose to use a simple technique called "NDF shading". It all...
متن کامل