Image Based Modeling and Rendering (PI: Malik) - PowerPoint PPT Presentation

About This Presentation
Title:

Image Based Modeling and Rendering (PI: Malik)

Description:

Acquire photographs. Recover geometry (explicit or ... Photographs are not ... Multiple photographs of the same scene. Output. Geometric meshes of each ... – PowerPoint PPT presentation

Number of Views:100
Avg rating:3.0/5.0
Slides: 16
Provided by: jitendr9
Category:

less

Transcript and Presenter's Notes

Title: Image Based Modeling and Rendering (PI: Malik)


1
Image Based Modeling and Rendering (PI Malik)
  • Capture photorealistic models of real world
    scenes that can support interaction
  • Vary viewpoint
  • Vary lighting
  • Vary scene configuration
  • This objective was achieved by combining and
    inventing new techniques in computer vision and
    graphics.

2
Image-based Modeling and Rendering
  • 1st Generation---- vary viewpoint but not
    lighting
  • Acquire photographs
  • Recover geometry (explicit or implicit)
  • Texture map

3
Image-based Modeling and Rendering
  • Photographs are not Reflectance Maps !
  • 2nd Generation---- vary viewpoint and lighting
    for non-diffuse scenes
  • Recover geometry
  • Recover reflectance properties
  • Render using light transport simulation

Illumination
Radiance
Reflectance
4
Image-based Modeling
  • 2nd Generation---- vary viewpoint and lighting
  • Recover geometry reflectance properties
  • Render using light transport simulation or local
    shading

Original Lighting Viewpoint
Novel Lighting Viewpoint
5
Inverse Global Illumination (Yu et al, SIGGRAPH
99)
Reflectance Properties
Radiance Maps
Light Sources
Geometry
6
Real vs. Synthetic
7
Real vs. Synthetic
8
Key contributions in Inverse Global Illumination
  • A digital camera can undertake all the data
    acquisition tasks involved.
  • Both specular and high resolution diffuse
    reflectance properties can be recovered from
    photographs.
  • Reflectance recovery can re-render non-diffuse
    real scenes under novel illumination as well as
    from novel viewpoints.

9
Image-based Modeling and Rendering
  • 3rd Generation--Vary spatial configurations in
    addition to viewpoint and lighting

Novel Viewpoint
Novel Viewpoint Configuration
10
Framework
  • Input
  • Multiple range scans of a scene
  • Multiple photographs of the same scene
  • Output
  • Geometric meshes of each object in the scene
  • Registered texture maps for objects

11
Segmentation ResultsYu, Ferencz and Malik TVCG
01
12
Models of Individual Objects
13
Texture-Mapping and Object Manipulation
14
Transitions
  • Façade has inspired several image-based modeling
    packages including Canoma from Metacreations.
  • Technique for registering range and radiance
    images being used by Cyra, a leading time of
    flight laser scanner company.

15
Continuing Challenges
  • Finding correspondences between widely separated
    camera views automatically
  • Models of reflectance and texture for natural
    materials and objects
Write a Comment
User Comments (0)
About PowerShow.com