1. Frame Buffer Applications
Fall 2018 CS480/680

2. Previously, What are frame buffers?
Different types of buffers stored in frame buffer
Color buffer
Depth buffer
Stencil buffer
How to create and use multiple frame buffers
Use of color buffer to develop post processing shaders

3. Other Buffers in Frame Buffers
Currently, we’ve used color buffers to create post-processing shaders
Frame buffers can be used for a lot more than just modifying the screen pixels
There are still two other buffers that we can use
Depth buffer for storing position
Stencil buffer for masking
What can we do by using the depth buffer?

4. Shadow Mapping

5. The global lighting problem
Our current lighting algorithm only works with lighting individual objects (even then, it doesn’t do a good job)
Other object’s don’t know if they are being occluded or not

6. Solution: shadow mapping
We can tell if a pixel is occluded or not by drawing an imaginary line from the light source to the fragment, and see which surface it hits first
We can use a depth map from the light source to the object to get that imaginary line AND the surface that it hits first
Only problem is, is that the line is in light space; how do we compare a world space coordinate to a light space coordinate?

7. Solution: shadow mapping
Matrix Transformations!
Transform the fragment position into light position, and compare the first hit object to the current fragment
If the first hit object is closer than the current fragment, it's in shadow

8. Learning By Example

9. A still imperfect method
While shadow mapping is good, we still make the assumption that the light is an infinitely small point
Real lights have a size to them, which causes the soft shadows we see cast by lamps and computer monitors

10. Deferred Rendering

11. Optimizing the Rendering Lights
On the topic of light rendering, you might notice rendering a lot of lights gets slow very fast
What we are currently using a technique called Forward Rendering
We waste a lot of time calculating lights that are too far to be lit by some lights, or are out of the area of influence (i.e. spotlights)
We also do light calculations for pixels that won’t even be seen (overdraw)
When doing fragment lighting, we calculate the specular and diffuse products of all objects for all lights in all pixels, even if they are not seen

12. Deferred Rendering
Solution: don’t do any lighting calculations until after the whole scene is rendered
Wait what
We know all the lighting information we need:
Texture
Normals
Fragment Position (for calculating L and V)
Material info (specular power, diffuse power, etc.)
We pack them into frame buffers, and do the calculation afterwards

13. Optimization 2: light volumes
We solved the issue of overdraw, but we are still calculating all the lighting for all pixels for each scene
Solution: we mask out areas we know the light will affect using a model
We can exploit the efficiency of the rasterizer to only calculate lighting on the pixels that will be affected by the light
We use primitive shapes to mask out these areas:
Spheres for point lights
Cones for spot lights
A fullscreen quad for directional lights (they have no attenuation or area of influence)

14. Learning By Example

15. Example in Industry: Grand Theft Auto V

16. Questions?

17. Additional Resources