Přednáška Fotorealistická syntéza obrazu (10.-17. 12. 2008)

Prezentace k přednášce (>100 stran v češtině)

Images courtesy of Henrik Wann Jensen, James Arvo, Eric Veach, Leonidas Guibas, et al.

Local reflectance models

Metal teapot Diffuse teapot Glass teapot Glossy teapot

Metal surfaces rendered using different light models:

Gold, Phong Gold, Hall Gold, Cook-Torrance Copper, Cook-Torrance

Sample materials from Cornell BRDF database:

Automotive paint Spray paint House paint
Ceramics Human skin  


Images generated by PoV ray tracer:

Manhattan The Lovers Pearl Harbor Warm Up
Marbles Office Kitchen The Classroom


Flat shading Gouraud Gouraud, Yellow light
rad-flat rad-gou rad-gou-yellow

Combined heuristics

Light source sampling BRDF sampling Combined sampling

Global illumination


Photon mapping Photon mapping
Sun is the only light source


Photon mapping Photon mapping

All images have been rendered at 1024x768 with 4 samples per pixel on a Dual P3 800MHz Linux-PC.

Ray tracing
1.5 sec
+ soft shadows
7 sec
+ caustics
12 sec (2+10)
+ global illumination
15 sec (3+12)
caustics photons 50k
global photons 200k
Path tracing reference

Combined heuristics (Bidirectional Path-tracing)

Path-tracing Combined heuristics

Individual components (light paths):

Cornell box

Spectral Monte-Carlo
Photon mapping + glass sphere Transcluent box
Fractal walls
1.6mil. elements
caustics: 50k photons
global: 200k photons
14min (2xPII/400)
Fractal wall & spheres
8min (2xPIII/800)
Diffuse box
4 samples/pixel
9min (Pentium/100)
Glossy box
16 samples/pixel
50min (Pentium/100)
Glossy box reference
1000 samples/pixel

Metropolis Light Transport

Caustics details:

Bidirectional Path-Tracing Metropolis Light Transport

[Public seminars]  [J. Pelikán's home page]  [CGG server]

Valid XHTML 1.1 Copyright: text, slides (C) 2006-2008 J.Pelikán
Copyright: images (C) Henrik Wann Jensen, James Arvo, Eric Veach, Leonidas Guibas et al.
last change: $Date: 2013-11-22 23:48:14 +0100 (Fri, 22 Nov 2013) $
Send comments/requests to pepca.at.cgg.mff.cuni.cz