Docstoc

t_intro

Document Sample
t_intro Powered By Docstoc
					    Multiresolution Meshes in Surface Modeling
I
N
T
R                     Leila De Floriani
O
D               DISI - University of Genova (Italy)
U                               and
C                           UMIACS
T
I
O
N

                         Joint work with
              Paola Magillo and Enrico Puppo
                  (DISI - University of Genova)




                                                      1
                                                              Outline
       Introduction: motivations and background
       LOD Models: layered versus multiresolution models
I
N
       A general framework for multiresolution models: the Multi-
T       Triangulation
R
O
       Classification of Multiresolution Models: tree-like and
D
U       evolutionary models
C
T      Tree-like Models: nested models based on regular or irregular
I
O
        subdivisions
N
       Evolutionary Models: construction strategies and encoding
        structures
       A taxonomy for multiresolution models



                                                                        2
                                                            Motivations
       High complexity of 3D scenes

         automatic acquisition of the surface of solid objects
I
             range scanners -- 3D scanners
N
T            resolution supported: ~ 10 facets / mm2
R
O        standard solid modeling tools (CAD)
D
U            complex 3D object defined by # faces > 100K
C
T        digital terrain models
I
O            millions of faces
N
         tesselation of implicit surfaces




                                                                          3
                                                              ...Motivations...
       3D graphics is a limited, valuable resource

          graphics throughput of low level ws / pc : ~100K faces/sec
I
N
T         interactivity requires multiple frames per second
R
O         web graphics (VRML) needs trasmission of data on low
D
           bandwidth networks
U
C             scientific users: 300-500KB/sec local, 10 - 100KB/sec remote
T
I             commercial/home users: 56Kb/sec                      (in Italy)
O
N          graphics file size:
              24 byte/vertex    ( if binary, >> if ascii )




                                                                                  4
                                   Reducing Graphics Costs
       Visualization Stage:
          culling back faces
I         view frustum culling
N         visibility culling
T
R
O      Modeling Stage:
D
U         tessellate surfaces with triangle meshes
C
          simplify meshes
T
I         construct a LOD model
O
N




                                                             5
                                            View Frustum Culling
       an example of view frustum culling
        (images by SGI, OpenGL Optimizer)
I
N
T
R
O
D                                       view
U                                     frustum
C
T
I
O
N


                  culling OFF                      culling ON




                                                                   6
                                              Visibility Culling

       an example of occlusion/visibility culling
        (images by SGI, OpenGL Optimizer)
I
N
T
R
O
D
U
C
T
I
O
N




                                                                   7
                                     Level of Detail:
        Approximating surfaces with triangle meshes

       Assumption:
I
N
T
          accuracy of the approximation is proportional to the number of
R          triangles
O
D
U      Objective:
C
T         always produce the simplest mesh that satisfies the accuracy
I
O
           required by the application
N




                                                                            8
                                       …Approximating Surfaces...
       On-the-fly simplification: extract from raw data a mesh of
        minimal size whose accuracy is sufficient for application needs
I             only raw data and the simplified mesh are stored
N
T
              simplification is usually an expensive task
R
O
D
U
C
T
I
O
N




                                                                          9
                                       …Approximating Surfaces...

       LOD / multiresolution model: build a model off-line that
        encompasses many different representations and that can be
I       queried efficiently
N
              more expensive in terms of space
T
R             more efficient: support to real-time operations
O
D
U
C
T
                                  LOD model
I
O
N




                                                                     10

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:0
posted:2/15/2012
language:
pages:10