Combining Positions and Normals for Precise 3D Reconstruction

1
Combining Positions and Normals for Precise 3D
Reconstruction

• Speaker ???
• Advisor I-Chen Lin

2
Outline

• Introduction
• Related work
• Future work

3
Introduction

• Shape acquisition

4
Active Stereo

• High quality positions
• But high frequency noise
• Stereo match triangulation
• Structured Light System

5
Passive approach

• Avoiding the correspondence issue.
• High frequency details
• But Low frequency in nature
• Photometric stereo
• Shape from shading

6
Combine Positions and Normal

• Rendering demands high quality normal.
• Geometry processing requires high quality
  positions.
• Measure positions and normal independently
• Combine them into 3D Model better.

7
Related work

• Active stereo
• Calibration of a Structured Light System A
  Projective Approach Huynh CVPR 97
• Spacetime Stereo Recovery for Dynamic Scenes Li
  Zhang CVPR 2003

8
Related work

• Passive Approach
• Example-based photometric stereo shape
  reconstruction Hertzmann CVPR 2005
• Recovering Shape and Reflectance Model of
  No-lambertian Objects from Multiple Views
• Yu CVPR 2004

9
Related work

• Combine
• Efficiently Combining Positions and Normals for
  Precise 3D Geometry
• Nehab SIGGRAPHIC 2005

10
Future work

• Contribution
• A Precise 3D Reconstruction
• Can fit hole
• For general object ex glassy object

11
Future work

• Positions (already did)
• Coded Structured light system
• Projective Calibration Huynh CVPR 97
• Coded light INOKUCHI CVPR 84
• An additional periodical pattern is projected

12
Future work

• Normals
• Shape from shading (using BRDF model)
• Combine
• Challenge (how to better then others?)
• Yu CVPR 2004
• brdf model, multi-material, initial step etc
• Nehab SIGGRAPHIC 2005
• weight, material property,
• Using one cost function to minimize err
• and iterative optimization

13
Q A

• Thanks a lot.