Inverse Rendering: From Concept to Applications

Abstract

Inverse problems are usually of extreme complexity and are an important research topic for the graphics community due to their wide applicability. Those problems are considered for which the input data is captured from reality with a camera, so it is possible to extract information about the scene illumination, reflectance properties, or geometry. Unfortunately, sometimes standard techniques fail due to practical issues: uncertainty in acquisition methods, dynamic behaviors, complexity of the scene and approximate geometrical or lighting model. This course not only aims at the presentation of the fundamental principles behind inverse rendering problems, but also presents some practical considerations that arise in some applications, explains the new problems introduced and shows adequate solutions. In the introduction, a review is given of the course objectives, the context and some of the prerequisites, e.g. a general definition of global illumination. The first part of the course gives a definition of inverse rendering, and presents a selection of inverse rendering solutions present in the literature. The second part of the course, describes practical methods that can be used in specific applications of inverse rendering. Retrieving reflectance and illumination properties of an existing site is of interest in areas as post-production, architectural and urban planning, and interior lighting design. Modeling geometric properties of an object based on illumination output can be used in the design of reflectors in the car industry or urban lighting design. The course is given as a lecture and is supported by slides and videos. Material created by the course organizers and coming from other sources is used to illustrate the ideas presented in the course. At the end of the course, attendees will have a good understanding of inverse rendering problems, and be able to select a method from the literature to apply to a specific application.