Mesh colorization presentation

22
Mesh Colorization Author: George Leifman and Ayellet Tal EUROGRAPHICS 2012, May in Italy Speaker: Chung-Yuan Lee Yuan-Ze University

description

Mesh colorization presentationby Chung-Yuan Lee from Yuan-Ze University, Taiwan

Transcript of Mesh colorization presentation

Page 1: Mesh colorization presentation

Mesh Colorization

Author: George Leifman and Ayellet TalEUROGRAPHICS 2012, May in Italy

Speaker: Chung-Yuan LeeYuan-Ze University

Page 2: Mesh colorization presentation

OutlineIntroductionProposed method

◦Basic algorithm Vertex similarity Color propagation

◦Feature line fieldResultReference

Page 3: Mesh colorization presentation

Introduction

A novel algorithm for colorization of meshes.

Page 4: Mesh colorization presentation

Basic algorithmVertex similarity◦Step1. Vertex descriptor◦Use spin image to describe vertex

spreading.

Page 5: Mesh colorization presentation

𝑆𝑝𝑖𝑛 𝑖𝑚𝑎𝑔𝑒[1]2D histograms

Page 6: Mesh colorization presentation

Use a relatively small support region, by setting α= β=8

Page 7: Mesh colorization presentation

Basic algorithmStep2. Similarity measure

Model the difference between two histograms

where

Page 8: Mesh colorization presentation
Page 9: Mesh colorization presentation

σ is a constant standard deviation for the Gaussian filter ϕ

The notation denotes half size down-sampling

Page 10: Mesh colorization presentation

k(α) is the norm, making the diffusion distance a true metric.

This method use L = 3 and σ = 0.5

Page 11: Mesh colorization presentation

Basic algorithmStep3. Edge assignment

means similarity between the vertices I and j

MaxD is the maximal distance between all pairs of neighboring vertices of the mesh

The "amount of color" that can flow between and

13

12

1

Page 12: Mesh colorization presentation

Basic algorithmColor propagationFor each color channel C in the YUV color space,the method minimize the following cost function.

This equation constraint that two neighboring vertices should have similar colors if their geometry is similar.

Page 13: Mesh colorization presentation
Page 14: Mesh colorization presentation

The result might include color bleeding near the boundaries !

Page 15: Mesh colorization presentation

Feature line fieldDefinitionFor neighboring points that belong

to the same region, the directions of the associated vectors should have similar orientations.

And vice versa!

Page 16: Mesh colorization presentation

Feature line fieldComputationThis method describe the

computation for the case of valleys.

Ridge −Valley  Lines  on  Meshes   via   Implicit   Surface  Fitting  [2 ]

Page 17: Mesh colorization presentation

Surrounding face

Feature line field direction lies on the line perpendicular to the edge adjacent to it

The average of all the field vectors computed for each face independently

Solve the Laplace’s equation

sum of cotangents of the angles opposite the edge (v, ) in the triangles sharing this edge

Page 18: Mesh colorization presentation

Similarity measure adjustment

Page 19: Mesh colorization presentation

Case1:

Case2:

ԑ is set to half of the shortest edge of the mesh

𝑣 𝑖 𝑣 𝑗

Page 20: Mesh colorization presentation

Result

Page 21: Mesh colorization presentation
Page 22: Mesh colorization presentation

Reference[1] “Spin Images”

◦Georgios Papadimitriou [2]Ridge-Valley Lines on Meshes

via Implicit Surface Fitting◦Yutaka Ohtake , Alexander Belyaev ◦Hans-Peter Seidel 2