Problèmes inverses pour la 3D

Objectives

This course aims to give an overview of the methods of photographic 3D reconstruction. The problem is the following: from one or more photographs, how to obtain a 3D model of a scene, i.e. its relief and color? We show that the 3D reconstruction methods can be reduced to the resolution of inverse problems. We introduce the main tools for solving such problems. A certain number of sessions of this course will take place in the form of practical work or laboratory work, which will be inspired by the work carried out in the REVA team of IRIT.

Description

This UE is articulated around 3 parts:
PART 1: GEOMETRY: Generalities on photography, 3D as an inverse problem. Introduction to Meshroom, Structure-From-Motion, reminder of optimization tools, beam fitting for SfM
PART 2: Shape-From-Shading: Lecture on light, Photometry, BRDF, lighting representation, underlying inverse problems, Fast marching, depth sensors
PART 3: Stereophotometry

Bibliography

- R. Szeliski, "Computer Vision: Algorithms and Applications", 643 pages, 2010, Springer.
- D. Forsyth, J. Ponce, "Computer Vision: a Modern Approach", 792 pages, 2011, Pearson

Pre-requisites

Optimization, image processing