PDF (6.5M)

6
Procedural Isotropic Stochastic Textures by Example Supplemental Material Ares Lagae 1 , Peter Vangorp, Toon Lenaerts, Philip Dutr´ e Department of Computer Science, Katholieke Universiteit Leuven Abstract This document contains additional material for the submission Procedural Isotropic Stochastic Textures by Example. Additional Results Figure 1 and figure 2 are extended versions of the corre- sponding figures in the paper. These figures also contain con- text photographs and additional results. Comparison with the Method of Heeger and Bergen In the paper we state that the method of Heeger and Bergen and our method achieve a similar synthesis quality. This is il- lustrated in figure 3. Integration into Production Rendering Software In the paper we state that our procedural textures can eas- ily be integrated into production rendering software. This is illustrated in figure 4, which shows a screenshot of our method integrated into Autodesk Maya. Implementation as a GPU Shader In the paper we state that our procedural textures can easily be implemented as a GPU shader. This is illustrated in figure 5, which shows a screenshot of a procedural texture implemented as a GPU shader. Histogram Fitting Figure 6 shows a clearer version of the histogram and the fitted histogram (the splines) of the corresponding figure in the paper. Email address: [email protected] (Ares Lagae) 1 Ares Lagae is visiting at REVES/INRIA Sophia-Antipolis in 2009-2010. Preprint submitted to Computers & Graphics (Special issue on Procedural Methods in Computer Graphics) May 24, 2010

Transcript of PDF (6.5M)

Page 1: PDF (6.5M)

Procedural Isotropic Stochastic Textures by ExampleSupplemental Material

Ares Lagae1, Peter Vangorp, Toon Lenaerts, Philip Dutre

Department of Computer Science, Katholieke Universiteit Leuven

Abstract

This document contains additional material for the submission Procedural Isotropic Stochastic Textures by Example.

Additional Results

Figure 1 and figure 2 are extended versions of the corre-sponding figures in the paper. These figures also contain con-text photographs and additional results.

Comparison with the Method of Heeger and Bergen

In the paper we state that the method of Heeger and Bergenand our method achieve a similar synthesis quality. This is il-lustrated in figure 3.

Integration into Production Rendering Software

In the paper we state that our procedural textures can eas-ily be integrated into production rendering software. Thisisillustrated in figure 4, which shows a screenshot of our methodintegrated intoAutodesk Maya.

Implementation as a GPU Shader

In the paper we state that our procedural textures can easilybe implemented as a GPU shader. This is illustrated in figure 5,which shows a screenshot of a procedural texture implementedas a GPU shader.

Histogram Fitting

Figure 6 shows a clearer version of the histogram and thefitted histogram (the splines) of the corresponding figure inthepaper.

Email address: [email protected] (Ares Lagae)1Ares Lagae is visiting at REVES/INRIA Sophia-Antipolis in 2009-2010.

Preprint submitted to Computers & Graphics (Special issue on Procedural Methods in Computer Graphics) May 24, 2010

Page 2: PDF (6.5M)

(a) Vinyl flooring. (b) Blue towel.∗

(c) Rock salt lamp.∗ (d) Bluestone doorstep.∗

(e) Ceramic shower tile. (f) Blanket. ∗

(g) Octagonal table. (h) Living room tile.

(i) Jeans vest.∗ (j) Rust.

(k) Clinker. ∗ (l) Granite (Rose Sarde).∗

(m) Granite (Bethel White).∗ (n) Granite (Shivakasi Yellow).∗

Figure 1: Several examples of our method for texture synthesisby example for isotropic stochastic procedural textures. Each example shows a photograph withcontext, a photograph of a texture and a procedural texture automatically generated from the photograph using our method.The procedural textures in the examplesmarked with an asterisk∗ were constructed using histogram matching.

2

Page 3: PDF (6.5M)

(a) Purple towel.∗ (b) Brick wall.

(c) Limestone brick.∗ (d) Ceramic tile.

(e) Concrete ceiling.∗ (f) Stone tile.∗

(g) Granite (Jasberg).∗ (h) Granite (Rustenbuild).∗

(i) Granite (Giallo Veneziano).∗ (j) Marble (Azul Valverde).∗

(k) Marble (Bianco Royal).∗ (l) Granite (Mountain Pink).∗

(m) Granite (Porrino).∗ (n) Human skin.

Figure 2: Several examples of our method for texture synthesisby example for isotropic stochastic procedural textures. Each example shows a photograph withcontext, a photograph of a texture and a procedural texture automatically generated from the photograph using our method.The procedural textures in the examplesmarked with an asterisk∗ were constructed using histogram matching.

3

Page 4: PDF (6.5M)

Exemplar Heeger and Bergen Ours Exemplar Heeger and Bergen Ours

(a) (b)

(c) (d)

(e) (f)

Figure 3: A comparison of our method to the method of Heeger and Bergen. Each example shows a photograph of a texture, an image texture synthesized from thephotograph using the method of Heeger and Bergen, and a procedural texture automatically generated from the photograph using our method. The results of bothmethods are similar, but the method of Heeger and Bergen cannot be formulated as a procedural texture.

4

Page 5: PDF (6.5M)

Figure 4: Integration into production rendering software.A screenshot of our method integrated intoAutodesk Maya.

Figure 5: Implementation as a GPU shader. A screenshot of a procedural texture implemented as a GPU shader.

5

Page 6: PDF (6.5M)

0

0.2

0.4

0.6

0.8

1

0 50 100 150 200 250

coun

t

bin

Histogram

(a)

0

0.2

0.4

0.6

0.8

1

0 50 100 150 200 250

coun

t

bin

Histogram

(b)

Figure 6: Histogram fitting. (a) Histogram of the exemplar. (b) Fitted histogram of the exemplar.

6