47
Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun [email protected] [email protected] 2003 . 4.29

Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun [email protected]@cc.gatech.edu [email protected]@cc.gatech.edu 2003. 4.29

Embed Size (px)

Citation preview

Page 1: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Poisson Image Editing& Terrain Synthesis

Howard Zhou Jie [email protected] [email protected]

2003 . 4.29

Page 2: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Table of Contents

1. Introduction / motivation2. Poisson Image Editing3. Terrain Synthesis (Texture based methods)

4. Future work5. Conclusion

Page 3: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Table of Contents

1. Introduction / motivation2. Poisson Image Editing3. Terrain Synthesis (Texture based methods)

4. Future work5. Conclusion

Page 4: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Introduction / motivation

Poisson Image Editing Seamless

Texture based terrain synthesis Current method based on fractals Very limited control Terrain style adjusted by parameter tuning What if the user draws a rough sketch and

supply a height map and says: “I want this to be like this”

Page 5: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Poisson Image Editing

Review

Page 6: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Our implmentation

Matlab Sparse matrix PDF solver Use conjugate gradient solver

supplied by Matlab Can be faster if …

Page 7: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Seamless insertion

Page 8: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Inserting objects with holes

Page 9: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Inserting transparent objects

Page 10: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Texture flattening

Result directly related to Edge detection result

Page 11: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Local illumination changes

alpha = 0.05beta = 0.2

alpha = 0.05beta = 0.4

Page 12: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Seamless tiling

Good when seam is not significant Often needs to increase the contrast of

the result but don’t an automatic way, maybe use

histogram of the original image

Page 13: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Seamless tiling

Good when the seam is not significant

Page 14: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Seamless tiling

Show some more

Page 15: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Seamless tiling

Page 16: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Seamless tiling

Page 17: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Seamless tiling

Contrast can be globally fixed

But how?

Page 18: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Seamless tiling

Seams not good

Cannot be fixed

Page 19: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Table of Contents

1. Introduction / motivation2. Poisson Image Editing3. Terrain Synthesis (Texture based methods)

4. Future work5. Conclusion

Page 20: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Previous approachTexture based terrain synthesis

Current method based on fractals Very limited control Terrain style adjusted by parameter

tuning What if the user draws a rough

sketch and supply a height map and says: “I want this to be like this”

Page 21: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Texture based terrain synthesis

1. Image analogy2. Texture synthesis on laplacian +

piecewise seamless tiling3. Graph cut / seamless tiling4. Separating the details

Page 22: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Data: height map

Page 23: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Display height map

Page 24: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Image analogy

                        

       

:

                        

       

::

                             

:

                          

       

A A' B B'

A. Hertzmann, C. Jacobs, N. Oliver, B. Curless, D. Salesin.SIGGRAPH 2001

Page 25: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Texture by number

Page 27: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

How do we get (A) automatically

Blurring (filtering) Texture flattening using edge

detection result or contour

Page 28: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Image analogy + (texture flattening + blurring)

Page 29: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Laplacian Synthesis

Regard laplacian as a particular texture

Texture synthesisIntegrate

Page 30: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Results

Page 31: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Terrain

Page 32: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Terrain

Page 33: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Problems & possible solutions

Depend on the boundary conditionsUse the boundary attached to the

Laplacian

There is only one unique solution of this linear system

Lost the power of Poisson editingShould use a non-conservative

gradient field

Page 34: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Graph cut + seamless tiling

Page 35: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Laplacian removing boundary(since the boundary is known)

Page 36: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Image smoothing edge (1 D)

Page 37: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Using Poisson Solver

Page 38: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Terrain Analysis

The detail of the terrain differs at different altitude

Terrain = f ( altitude )Altitude = g ( style )

Page 39: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Example: Terrain map

Page 40: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Low Frequency - Altitude

Page 41: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

High Frequency – as a function of Altitude

Page 42: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Proposed Algorithm

Use “Copy & Paste” methods to generate an altitude map

Add high frequency probabilistically as indexed by the altitude map

Graph cuts/Image Quilting to make it seamless

Page 43: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Table of Contents

1. Introduction (motivation)2. Re-illumination3. Changing viewpoint4. Future work5. Conclusion

Page 44: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Future Work

Other texture methods (Graph cut, stocastic?)

Stylized map generation from real map Real map from stylized map

Page 45: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Map vs. terrain

Page 46: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Conclusion

Implemented poisson image editingTried several texture based terrain

synthesis methodsLots to be done!

Page 47: Poisson Image Editing & Terrain Synthesis Howard Zhou Jie Sun howardz@cc.gatech.eduhowardz@cc.gatech.edu sun@cc.gatech.edusun@cc.gatech.edu 2003. 4.29

Questions ?