Adaptive Shadow Maps - PowerPoint PPT Presentation

1 / 11
About This Presentation
Title:

Adaptive Shadow Maps

Description:

ASM queries using depth culling. Store most recently queried leaf node in cache. SSE/SSE2 ... Frustum culling for each cell in top level of hierarchy ... – PowerPoint PPT presentation

Number of Views:43
Avg rating:3.0/5.0
Slides: 12
Provided by: dongkyu
Category:

less

Transcript and Presenter's Notes

Title: Adaptive Shadow Maps


1
Adaptive Shadow Maps
  • R. Frenando, S. Fernandez,
  • K. Bala, Donald P. Greenberg

2
Introduction
  • Shadow map
  • A depth image generated from the light
  • Fast but can introduce aliasing
  • Adaptive shadow map
  • Use a hierarchical grid structure for depth
  • View-driven
  • Progressive improvement

3
(No Transcript)
4
Previous Work
  • To remove aliasing
  • Percentage-closer filtering
  • Light buffer for non-interactive ray tracer
  • Deep shadow map using jittered samples
  • But do not deal with aliasing caused by
    insufficient shadow map resolution

5
Adaptive Shadow Map
  • Idea
  • Only regions that contain shadow boundaries need
    to be samples densely
  • Function
  • Input a set of transformed eye view points
  • Output true if a point is lit, false otherwise

6
Adaptive Shadow Map
  • Organization
  • Tree structure
  • Each node has a shadow map and a partitioning of
    shadow map into a fixed number of cells
  • Each cell may contain another tree node

7
Adaptive Shadow Map
  • Creating nodes
  • Use a cost-benefit metric to determine which
    cells to be assigned new nodes
  • Quantifying benefit counting number of
    transformed eye pixels and whose shadow map
    resolution is lower than the eye view resolution
  • ? mip-mapping capability of graphics H/W

8
Adaptive Shadow Map
  • Removing nodes
  • A particular nodes memory need to be reclaimed
  • Use LRU on all nodes that were not used in last
    frame
  • If there are no such nodes, remove existing node
    only if new node has more benefit that existing
    one

9
Implementation
  • Mip-mapping
  • To determine when resolution mismatches occur
  • Graphics H/W tells interpolated mip-map level
  • Combining ID and depth comparison
  • First, ID test
  • If it fails, do the conventional depth test

10
Optimization
  • ASM queries using depth culling
  • Store most recently queried leaf node in cache
  • SSE/SSE2
  • Frustum culling for each cell in top level of
    hierarchy

11
Conclusion
  • Adaptive shadow map
  • Adaptive subdivision to address aliasing
  • View-driven
  • Progressive
  • Confined to a use-specifiable memory limit
  • Compatible with conventional shadow map
Write a Comment
User Comments (0)
About PowerShow.com