The Realistic Lighting Using Environment Maps

Rendering computer-generated (CG) objects into real photographs is one of the most important problems in VR and AR applications. In general, such work needs recovering the real-world information, which includes camera positions and parameters, shapes, material properties, and motion of real objects.The illumination coherence is one of the key issues involved. To achieve this goal, current available techniques are labor intensive. A common technique is to manually recover the position of the light source with the help of a reference object, and to instantiate a virtual light of equal color and intensity for each real source to illuminate the CG objects.A new method, which illuminates the environment with environment maps, is proposed in this paper. We store the illumination in environment maps (light probe image) to simulate the true lighting environment. Then an illumination model based on environment maps was proposed. Meanwhile, a new fast prefiltering method for environment maps using spherical harmonics is discussed in this thesis.In Chapter 1, we first introduce the basic concepts and primary characteristics of VR systems. We also introduce AR which emerged these years briefly. Then we review the history of the use of environment maps in image-based rendering. The central topic, that is a new approach to obtain illumination coherence, is addressed in the end.In Chapter 2, we introduce the light probe image, which stores the true illumination information in environment maps. The basic principle, which includes the mapping method and pixel encoding was introduced. Meanwhile, we present a new practical approach to obtain the light probe image.In Chapter 3, we proposed an illumination model based on environment maps, which is the principle of lighting the CG objects. We illustrate the lighting computation based on photometry. The result of our experiment is also presented.In Chapter 4, we discuss a fast prefiltering method for environment maps based on spherical harmonics, which is the analogue on the sphere to the Fourier basis on the line or circle. The light probe image is a 2D plenoptic function and can be approximated by a serious of spherical harmonics. We present the implementation of the algorithm, which include the data structures and the description of some procedures. The prefiltered environment maps are also given as the result of our approach.Finally, we draw the general conclusion of this thesis in the last chapter. Future work plan on this specific topic is also included.