Animated Billboard Clouds for Foliage Simplification
Authors
Carlsen, Jeppe ; Schmid, Jakob ; Bennett-Therkildsen, Anders
Term
4. term
Education
Publication year
2007
Abstract
Inden for computergrafik bruges ofte billboard clouds (samlinger af flade, teksturerede billeder) til at erstatte kompleks 3D-løv i træer. Et kendt svagpunkt ved eksisterende metoder er, at de primært er udviklet til statiske modeller og derfor ikke forenkler animerede modeller særlig godt. Denne afhandling fokuserer på forenkling af animerede polygonale træmodeller og udvider teknikken til det, vi kalder animerede billboard clouds. Vi præsenterer to løsninger, der forenkler træers animation, samtidig med at udseendet bevares så tro mod originalen som muligt. Løsningerne bygger på vores egen implementering af Lacewell et al.s billboard-cloud-algoritme, som er udviklet til forenkling af løv i statiske træer. For at vurdere den visuelle lighed definerer vi fejlmål, der objektivt måler kvaliteten af forenklingen. Vi har udviklet en prototype af den ene løsning og tester og evaluerer den sammen med vores implementering af den valgte billboard-cloud-algoritme.
In computer graphics, billboard clouds (groups of flat, textured images) are often used to replace complex 3D tree foliage. A recognized weakness of existing methods is that they mainly target static models and therefore do not simplify animated models well. This thesis focuses on simplifying animated polygonal tree models and extends the technique to what we call animated billboard clouds. We present two solutions that simplify the animation of tree foliage while aiming to preserve the original appearance. Our solutions build on our own implementation of the billboard cloud algorithm by Lacewell et al., designed for simplifying foliage in static trees. To assess visual fidelity, we define error metrics that objectively measure the quality of the simplification. We also develop a prototype of one solution and test and evaluate it together with our implementation of the chosen billboard cloud algorithm.
[This abstract was generated with the help of AI]
Documents