Final Project
Group Members
Nicolas Benitez
Michael Graeb
Proposal
For our final project, we will be attempting to create an image of a waterfront city similar to those pictured below.
The main technical goal of our project will be to investigate and implement a time of day-dependant lighting system that simulates the different types of environmental lighting found in outdoor areas. We hope to create a lighting system that will approximate the changes in spectra that occur at different times of day and produce convincing lighting and sky color. We hope that our lighting model will be allow us to reproduce the dramatic coloration that is seen on the reflective building surfaces in our referance images. In addition to helping reproduce our source images, such a lighting mechanism would be a very convenient resource to have when rendering any outdoor scene, since a time could be specified and appropriate lighting and sky coloration would be generated procedurally without needing to capture an environment map for the scene.
Other interesting aspects of the source images include the buildings, clouds, and water. We will be creating the scene geometry in Maya and exporting it to PBRT using the plugin featured at http://graphics.cs.ucf.edu/mayapbrt/ . If time permits, we would also like to pysically simulate the clouds and water, although our main focus will be the sky and lighting model.
The sky will be computed as a preprocess and stored, in a texture, for later rendering the scene. The steps that we will follow to generate this map are described in the paper by Haber, Magnor, and Seidel and can be summarized as follows:
- Compute the position of the sun based on time of day
- Create a series of "shells," representing the varying layers of atmosphere around the observer
- Further split the shells into "cells"
- Calculate direct illumination upon each cell, taking into account atmospheric effects such as refraction
- Calculate radiative heat transfer between cells to simulate light scattering throughout atmosphere
- Store illumination as environment map
For clouds, we would like to use the VolumeGrid DensityRegion included in PBRT in conjunction with either an existing data set, or a procedural technique for creating this type of data. We hope that this technique will work well with our light model since light will naturally penetrate the volume region more in thin areas of the cloud, resulting in the orange edges that can be seen in our first image.