# CS488 - Introduction to Computer Graphics - Lecture 23

1. End of course topics

# Natural Phenomena

## The General Idea

Many substances we want to include in our scenes have complex surface properties.

• Each one has its own unique empirical basis, theory(ies) and models.
• Scientists in the domain area provide the empirical basis and the theories. E.g. leaves of trees
• remote sensing
• botany
• biology
• biochemistry and biophysics
• For computer graphics we want nothing more than a model that describes how light interacts with the surface. The model should be
• simple => quick to compute
• able to generate as many surface variations as possible. E.g. leaves
• different species of tree
• spring, summer and fall leaves
• wet leaves and dry leaves
• leaves in a drought
• leaves of healthy and sick trees
• able to generate the variations from as few parameters as possible

How it's done ab initio

1. Build a model, often a simulation, as close to the underlying science as possible. E.g. colours of cloth
• all normally-used pigments
• how pigment particles can be distributed in the matrix of threads
• how light travels through the matrix and interacts with pigments
2. Generate a full range of results from the model by varying its parameters. E.g. cloth dying
• spectral BRDF of surface as a function of
1. pigment densities
2. cloth fibre
3. cloth weave
4. surface treatment, such as how the cloth is washed dried and ironed
3. Reduce the parameters
• sometimes by understanding the science
• e.g. Kubelka-Munk theory
• sometimes by brute force.
• e.g. singular value decomposition
4. Build a modelling interface that allows the modelling to "paint by numbers".

#### Paritioned Rendering

This is an example of a general technique called partitioned rendering.

1. In nature the full simulation is performed during the rendering process.
2. For computer graphics we partition the rendering into two parts
1. an off-line calculation that leaves us a simplified model with parameters
2. an on-line calculations tha renders from the model as if the simulation were being performed on-line.

Partitioned rendering is very common in computer graphics

• For example, texture mapping or bump mapping.

## Human Skin

Physical picture

• keratin
• melanin
• blood

Mix them by Kubelka-Munk

Get skin parametrized by three concentrations:

• all races
• all levels of sun exposure, but not freckles
• all different emotions, mostly bllod flow

Encapsulated in a spectral reflectance function. But there are more things to add, such as

• gloss from skin oil
• hair, either hair by hair or as an overal lcolourant
• asperity scattering