Solid Texture Synthesis from 2D Exemplars - PowerPoint PPT Presentation

1 / 11
About This Presentation
Title:

Solid Texture Synthesis from 2D Exemplars

Description:

Johannes Kopf, Chi-Wing Fu, Daniel Cohen-Or, Oliver Deussen, ... A 3D texture that describes solid material. Obviate the need for finding the parameterization of ... – PowerPoint PPT presentation

Number of Views:155
Avg rating:3.0/5.0
Slides: 12
Provided by: ytli
Category:

less

Transcript and Presenter's Notes

Title: Solid Texture Synthesis from 2D Exemplars


1
Solid Texture Synthesis from 2D Exemplars
  • Johannes Kopf, Chi-Wing Fu, Daniel Cohen-Or,
    Oliver Deussen, Dani Lischinski and Tien-Tsin Wong

From Siggraph 2007 Presented by Li, Ying Tsung
2
Outline
  • Introduction
  • Related work
  • Synthesis framework
  • Result

3
Introduction
  • What is solid texture?
  • A 3D texture that describes solid material
  • Obviate the need for finding the parameterization
    of the textured object
  • What is texture synthesis ?
  • Construct large image from small sample image

4
(No Transcript)
5
Related Work-1
  • Synthesis methods
  • Parametric based
  • Non-parametric based
  • Pixel based
  • Patch based
  • Optimization based

6
Related Work-2
  • Synthesis Solid texture
  • Stereological Techniques for Solid Textures
    Jagnow 2004

7
Synthesis framework - Overview
  • Optimization based method
  • Minimize a global texture energy function
  • Optimization step
  • Optimization phase
  • Search phase
  • Avoid local minimal by using Histogram matching
  • Re-weighting

8
Global energy function
  • sv a single voxel sv,i the
    vectorized neighborhoods of v in the
    slices orthogonal to I axis ev,I exemplar
    neighborhood closest to sv,I
  • Set r 0.8 to be more robust against outliers
    Kwatra 2005

9
Optimization step optimization phase
  • To avoid too blurry result
  • Treating the texels eu,I,v as points in a
    high-dimensional space
  • Cluster by Mean-Shift ? average only the dominant
    cluster Wexler 2007

10
Optimization step search phase
  • Search the nearest neighbor
  • Apply PCA to reduce dimensions ? preserve 95 of
    the variance
  • Using ANN approximate nearest neighbor library
    Mount and Arya 2006
  • Search neighbors only on sparse grid

11
Result
  • Time
  • 2.4 GHz CPU 1283 solid texture with 3
    channelsper texel takes anywhere between 10 90
    mins
  • Quality
  • With/Without feature map compare
  • Isotropic/Anisotropic different
  • Different with others compare
  • No blurring others have this artifact
  • Color smear independent synthesis of correlated
    color
Write a Comment
User Comments (0)
About PowerShow.com