Photon Maps

Oscar Ivarsson and Pär Eriksson

Global Illumination

using

 - Overview

 - Pass 1: Photon Map Construction

 - Pass 2: Rendering

 - Discussion

 - Future Work

Introduction

Overview

- View independent preprocessing

- More efficient than pure ray tracing

- Good for caustics

- Simulates diffuse reflections

Properties

Pass 1: Photon Map Construction

The Life of a Photon

"A photon checks in at a hotel and is asked if he needs any help with his luggage." 

"No, I'm travelling light."

Pass 1: Photon Map Construction

1. Emission

Pass 1: Photon Map Construction

2. Scattering

Pass 1: Photon Map Construction

3. Storage

Final image rendered using Monte Carlo ray tracing


Pass 2: Rendering

Rendering Equation 

- Rendering Equation can be subdivided

-Lr can be split into a sum of components and defined as:

-Direct illumination contribution

-Specular reflection contribution

- Caustics on diffuse & slightly glossy surfaces

- Soft indirect illumination

Pass 2: Rendering

Direct illumination (the first term)

Approximated:

- Radiance estimate obtained from the global photon map

Accurate:

- If all in shadow or not in shadow, assume the same.

- If mixed, send shadow rays

Two methods used

Pass 2: Rendering

What:

- Contribution via direct illumination by the light sources

Specular reflection (the second term)

What:

-Radiance reflected of specular and higly glossy surfaces

How:

- Evaluated with Monte Carlo ray tracing

- Importance sampling based on BRDF minimizes computation

Pass 2: Rendering

Pass 2: Rendering

Caustics (the third term)

What:

-Represent caustics on diffuse and slightly glossy surfaces.

How:

-Visualized directly using photons in the caustic photon map

-Number of photons must be high

Soft indirect illumination (the fourth term)

What:

-Incoming light which has been reflected diffusely at least once.

How:

-Since the incoming direction is stored with each photon we can integrate the information with any BRDF. (Accurate method) 

-In practice the approximation is limited to surfaces ranging from Lambertian to slightly glossy.

Pass 2: Rendering

Estimating Radiance using the Photon Map

What:

-Computing  radiance leaving an intersection point x.

How:

-Use the same approximation of dA as where a sphere is centered at is expanded until it contains and has radius r. "dA" is then approximated as "pi*r^2".

Pass 2: Rendering

Filtering the result

Why:

- Problems occur when the photon density is too low -> gives blurry results

How:

- By applying a cone-filter to the estimate the occurring blur can be compensated

-For each element in the sum, look up the contribution of the wight function

vs

Pass 2: Rendering

 - Efficient or time-consuming?

 - Easy implementation?

 - Smart emission?

Discussion

 - Store photons in participating media

 - Progressive photon map

 - Stochastic progressive photon map

Future Work

Made with Slides.com