Texture%20mapping

1
Texture mapping
2

• Texture
• Visual appearance of a real world surface

Texture mapping Adding texture detail to a
modeled surface
3
Why Texture Map

• Adds photorealism to 3D models
• Can assist in scan registration
• Can recover shape detail at higher resolution
  than range scans

4
History 1970s 1980s

• 1970s
• 1974 Original concept presented Catmull 74
• 1976 Reflection maps Blinn and Newell 76
• 1978 Bump mapping Blinn 78
• 1980s
• 1983 Texture mapping polygons in perspective
  Heckbert 83
• 1983 Filtering for antialiasing Williams 83
• 1984 Illumination mapping Miller and Hoffman
  84
• 1986 Environment maps Greene 1986
• 1986 Survey of texture mapping Heckbert 1986

5
History 1990s

• 1990s
• 1991 Interpolation for polygon texture mapping
  Heckbert 91
• 1992 Projective texture mapping Segal 92
• 1992 SGI RealityEngine Hardware texture
  mapping
• 1996 View dependent texture mapping Debevec et.
  96

6
Texture-modulated Quantities Modulation of
object surface properties Reflectance Color
(RGB), diffuse reflection coefficient kd
Specular reflection coefficient ks Opacity
(a) Normal vector N(P) N(P t N) or N
NdN Bump mapping or Normal mapping
Geometry P P dP Displacement mapping
Distant illumination Environment mapping,
Reflection mapping
7
Using photographs as textures
3D model
Debevec et al. 96, 98 Pulli et al.
97 Buehler et al. 01
8
Mechanisms of Reflection
source
incident direction
surface reflection
body reflection
surface

• Surface Reflection
• Specular Reflection
• Glossy Appearance
• Highlights
• Dominant for Metals

• Body Reflection
• Diffuse Reflection
• Matte Appearance
• Non-Homogeneous Medium
• Clay, paper, etc

Image Intensity Body Reflection Surface
Reflection
9
Rendering under novel illumination
3D model
BRDFs Bidirectional Reflection Distribution
Function
Lighting model
10
Diffuse Reflection and Lambertian BRDF
source intensity I
incident direction
normal
viewing direction
surface element

• Surface appears equally bright from ALL
  directions!
• (independent of )

albedo

• Lambertian BRDF is simply a constant

• Surface Radiance

source intensity

• Commonly used in Vision and Graphics!

11
Diffuse Reflection and Lambertian BRDF
12
Reflections

• Blinn 76 Reflection maps.
• Used to model an object that reflects its
  surroundings to the eye
• Texture environment (sphere, latitude/longitude
  map, cube)
• For each surface point, compute the polar
  coordinates of the reflected area with respect to
  the current viewpoint.
• Use those coordinates to map a 2D environment
  map.
• Filter accordingly

Environment map
Teapot with highlights
Blinn and Newel 1976. Texture and reflections in
computer generated images
13
Bump mapping

• Shade based technique.
• Displacement map
• Compute a new normal for each point P

Blinn 1978 Simulation of wrinkled surfaces
14
Visibility processing
Solved problem

• Very well studied problem in CG hidden surface
  removal
• z-buffer, back-face culling, painters algorithm,
  ray-casting
• Debevec 96, 98 z-buffer solution, 98 with
  polygon clipping and object space testing (if
  required)
• Rocchini 99 Ray casting approach accelerated
  with uniform grids

15
The mapping process
Screen space (s,t)
P. Heckbert 1986. Survey of Texture Mapping
16
Surface parameterization
P. Heckbert 1986. Survey of Texture Mapping
17
Texture aliasing

• A screen pixel may map to several texels
• Point sampling a high-frequency texture can cause
  aliasing

• The solution is to filter (avg) the texels at the
  cost of
• Expensive texel averaging that can reduce
  rendering rates
• Smoothing
• Reduce cost by prefiltering
• Filter shape
• What is a pixel ? A box, a rectangle, a circle?

P. Heckbert 1986. Survey of Texture Mapping
18
Filtering using pyramids

• Reduce the cost of filtering by creating a
  pyramid of pre-filtered images.
• Mip map Each pyramid level contains a 2n sized
  version of the image.
• Trilinear Interpolation is done across different
  levels of the pyramid. Total cost is constant.
• Supported by todays graphics hardware and APIs
  (OpenGL)

Lance Williams 1983 Pyramidal Parametrics
19
Filter comparison
Trilinear interpolation on a pyramid
Point sampling
Summed area tables
Elliptical Weighted Average
P. Heckbert 1986. Survey of Texture Mapping
20
Texture registration

• Camera calibration problem (very well studied)
• Find extrinsic (required) and intrinsic
  parameters (if not known)
• Good calibration important to avoid artifacts!
• Feature match
• Points, lines, other geometric feature
• Manual, automatic

No calibration req. Same camera used for range and color images Pulli 97 Manual Line matching Debevec 96 Point matching Rocchini 01 Automatic Rectangle matching Stamos 01 Silhouette Lensch 00
21
Texture reconstruction
Each reference image usually contributes to the
final result
22
Texture reconstruction - summary

• Problem how to find optimal weights for each
  camera
• Solutions proposed mostly consider viewpoint
  dependence
• Good for surfaces that are facing the camera.
• Good to capture specular highlights and other
  viewpoint dependent effects
• Not so good for surfaces at grazing angles where
  texture sampling density decreases
• What about non-view dependent texture mapping?
• Best solution domain and application specific.

23
Weighted average
Resolution weight
Reference image
Eye
Debevec 96 Debevec 98 Pulli 97 Buehler 01 View weight X X X X Normal weight X Field of view weight X X X Resolution X
24
Acquiring, Stitching and Blending Diffuse
Appearance Attributes on 3D Models

• C. Rocchini, P. Cignoni, C. Montani, R. Scopigno
• Istituto Scienza e Tecnologia dellInformazione

25
Synthetic images of our vase rendered a without
color, b with a naive mapping of the color
textures acquired by a commercial laser scanner
and c with our unshaded, locally registered and
cross-faded textures
26
Acquiring diffuse surface attributes
Rocchini et al. 01

• Create an albedo map

Enforce Lambertian surface assumption by removing
shadows and specular highlights

27
Photometric Stereo
Lambertian case
Image irradiance

• We can write this in matrix form

28
(No Transcript)
29
Acquisition of Surface Attributes

• define viewpoints
• capture multiple images from each viewpoint
• differ lighting between images

30
Un-shading of Images

• want illumination-invariant colors, not light
  direction dependent colors
• remove main shading effects
• direct shading
• cast shadows
• specular highlights
• but not inter-object reflections

31
Un-shading of Images
before un-shading
after un-shading
32
Fig. 6ab1. An example of two valid images, (a)
and (b) if we map image (b) on the mesh and
render a synthetic image (b1) using the same
viewpoint of image (a), then we can see how poor
and distorted the detail is in the right-most
side of the mesh obviously, mapping image (a) on
this mesh section can give a much better local
representation of the detail Fig. 7a,b. Iterative
local optimization of texture coverage in the
sample drawing, vertices are initially assigned
to three target images (represented by a hexagon,
a square and a circle). Then, we select a set of
frontier vertices (indicated by arrows) and
change their target images, obtaining
configuration (b), which now corresponds to a
local minimum. Frontier faces areindicated with
an F in (b) Fig. 8. An example of optimized
frontier face management. Left we have 1137
frontier faces out of a total of 10 600 in the
initial configuration right we have only 790
frontier faces after optimization equal to the
projection of its vertices on ik this face is
called internal if, conversely, the vertices
of f are linked to two (or even three) different
target images, then face f is
33
40cm tall ceramic vase complex painted surface 8
views required running time of 89sec
Three different views of the resulting vase mesh
(rendered without shading using a standard
OpenGL-based interactive renderer). The image on
the bottom is a re-lighted image, obtained with a
photo-realistic rendering software
34
Results

• 25cm tall statuette
• complex shape
• 14 views required
• running time of 62sec