Occlusion Culling - PowerPoint PPT Presentation

1 / 56
About This Presentation
Title:

Occlusion Culling

Description:

Visibility culling. Goal: ... Dynamically culling view volume of the observer, producing a superset of the visible scene data. ... – PowerPoint PPT presentation

Number of Views:160
Avg rating:3.0/5.0
Slides: 57
Provided by: kao73
Category:

less

Transcript and Presenter's Notes

Title: Occlusion Culling


1
Occlusion Culling
  • I4B06 ???
  • I4B12 ???

2
Indoor scenes
3
Viewing position change
4
Visibility culling
  • Goal
  • To bring the cost of rendering a large scene down
    to the complexity of the visible portion of the
    scene.
  • Visible set
  • Conservative visibility
  • We may classify an occluded object as visible but
    may never classify a visible object as occluded.
  • PVS potentially visible set
  • Hierarchical scene

5
Visibility culling techniques
  • View frustum culling
  • Back-face culling
  • Occlusion culling

6
Taxonomy of occlusion culling techniques
  • Point vs. Region
  • Image precision vs. Object precision
  • Cell-and-portal vs. Generic scenes

7
Additional criteria
  • Conservative vs. approximate
  • All vs. subset of occludes
  • Individual vs. fused occluders
  • Need of precomputation
  • Treatment of dynamic scenes

8
Aspect graph in 2D
  • Principle
  • Subdivide
  • Visibility space partition (VSP)
  • Visual event
  • The boundary of a VSP region
  • End points
  • Supporting lines
  • Separating lines

9
Many Aspect region
10
Cell-and-portal from-region visibility
  • Based on the characteristics of architectural
    scenes ? cells and portals
  • Key concepts
  • Conservative visibility
  • PVS
  • Densely occluded environments

11
Important Points
  • Subdivided into convex cells
  • Adjacency graph
  • View volume
  • Dynamically culling ? view volume of the
    observer, producing a superset of the visible
    scene data.

12
A cell is visible
  • It is in the View Volume
  • All cells along stab tree are in the view volume
  • Sightline within the view volume exists through
    portals

13
Cells and portals adjacency graph and stab tree
14
Point-based object-precision method
  • Point-based Cells and Portals technique based on
    from-region methods Luebke and Georges
  • No pre-computing for each cell
  • Simple and effective
  • Rendering the initial cell and its portals
  • Any remaining visible geometry has lie inside the
    projection of those portals

15
Cells and Portals
16
Large convex occluders
  • Using of temporal coherence by taking a subset
    of visual events
  • Separating and Supporting planes
  • The basic idea ? aspect graph
  • Linearized portion of the aspect graph ( as the
    user moves)
  • The silhouette edges and vertices of the relevant
    primitives

Coorg and Teller
17
Large convex occluders (Improved)
  • View-dependent set of occluders
  • Object hierarchy
  • Visibility test
  • Pre-selected convex occluders
  • Bounding boxes of the hierarchy
  • Edges of the occluder vertices of the occludee

18
Culling using shadow frusta
  • Concept
  • A viewer cannot see the occludee if it is inside
    the shadow .
  • Dynamically choosing a set of occluders
  • ? Producing shadow frusta
  • ? Culling the bounding boxes a hierarchy of
    objects
  • Scene Test top-down

Hudson et al.
19
Point-based image-precision techniques
  • Key concept
  • Subsequent objects can be culled away quickly by
    the already-filled parts of the images
  • Finite resolution
  • More robust than object-precision
  • Numerical precision problems

20
Point-based image-precision algorithm
  • Hierarchical Z-buffer
  • Hierarchical occlusion map
  • Ray casting
  • Approximate volumetric visibility
  • Directional discretized occluders
  • Occluder shadow footprints

21
Ray casting
  • The simplest forms
  • Never rendering an occluded part of the scene
  • High complexity

22
Directional discretized occluders
  • Both HZB and HOM
  • DDOs simple concept
  • Replace complex occluder with simpler object

23
Approximate volumetric visibility
  • Using volumetric representation
  • Density of geometry in the environment ? (
    probabilistic value )
  • The volume opacity between regions (related to
    density)

24
Occluder shadow footprints
  • 2.5D scene ( terrain or height field )
  • Using z-buffer
  • Wedges ( delimit the occluder shadow )
  • Footprint
  • An object is Hidden

25
Shadow Footprint ???
26
Generic from-region visibility
  • Region or view cell
  • Sampling view points
  • Unacceptable flicking artifacts
  • Needing Effective occluder
  • Occluder fusion

27
Umbra Virtual occluder
28
Conservative visibility preprocessing using
extended projections
  • An extension of image-precision methods
  • Using Projection
  • Extended projection
  • Allowing occluder-fusion

Durand et al.
29
Visible or Occluded
30
Virtual occluders
  • View-dependent convex object
  • Represent the aggregate occlusion for a given
    cell
  • Applying an Effective from-region culling
    technique
  • To Efficiently compute a PVS

Koltun et al.
31
Virtual occluders
32
Occluder fusion for urban walkthroughs
  • Observation
  • It is possible to compute a conservative
    approximation of the umbra from a set of discrete
    point samples placed on the view cells boundary.
  • Shrinking an occluder by ?

Wonka et al.
33
5 sample points in view cell
34
Occlusion Culling Algorithms
  • Select from the whole scene that we want to
    render, and the selection is limited to those
    portions of the scene that re not considered to
    contribute to the final image.
  • The z-buffer selects and renders only those
    objects which are visible

35
Occlusion culling is useful
36
General occlusion culling
  • 1 OcclusionCullingAlgorithm (G) 2
    ORempty3 for each object g in G4
      if(isOccluded(g,OR))5     Skip(g)6
      else7     Render(g)8     Update(OR,g)9
      end
  • 10 end

37
Occlusion Culling for Outdoor
  • Outdoor environments
  • Closed-in locations

38
Bounding Volumes
  • A simple geometrical object roughly representing
    the volume of a real objects geometry
  • The most suitable geometric objects for bounding
    volumes are spheres and boxes
  • ( The goal of using a combination of both is
    speed )
  • Node Volumes

39
Bounding box and bounding sphere
40
The View Frustum
41
View Occluders
42
Occluder
43
Visibility in Difference Bounding
44
Hierarchy Tree Example Octree
45
Hierarchy Tree Example Octree
46
Hierarchical Z-Buffer
  • Replace Z-buffer with a Z-pyramid
  • Lowest level full-resolution Z-buffer
  • Higher levels derive from lower level

47
Hierarchical Z-Buffer
  • A Z-pyramid represents depth buffer in
    hierarchical fashion
  • Depth at each pixel at a given level represents
    the max of the depth of the four corresponding
    pixels at next level
  • Z-pyramid constructs until the farthest one
    derived

48
Hierarchical Z-buffer
49
Hierarchical Z-Buffer
  • Organize scene into an octree
  • Before rendering the polygons associated with a
    node in the octree, test if the nodes bounding
    box is visible
  • Recursion

50
Hierarchical Z-Buffer
1 ProcessOctreeNode(OctreeNode N)2
if(isOccluded(NBV, ZP)) then return3 for each
primitive p in N4    tileInto(p, ZP)5 end6
for each child node C in N in front-to-back
order7    ProcessOctreeNode(C)8 end
51
Hierarchical Occlusion Maps
  • Two major differences from HZB
  • Separates occluders from occludees
  • Divide occlusion test into two part
  • an depth test and a overlap test

52
Hierarchical Occlusion Maps
Depth
Overlap
53
Hierarchical Occlusion Maps
  • Representation of projection for overlap test
    occlusion map
  • Generate by rendering occluders

54
Occlusion Maps
Occlusion Map
Rendered Image
55
(No Transcript)
56
END
Write a Comment
User Comments (0)
About PowerShow.com