Output-Sensitive Rendering of Detailed Animated Characters for Crowd Simulation

Abstract

High-quality, detailed animated characters are often represented as textured polygonal meshes. The problem with this technique is the high cost that involves rendering and animating each one of these characters. This problem has become a major limiting factor in crowd simulation. Since we want to render a huge number of characters in real-time, the purpose of this thesis is therefore to study the current existing approaches in crowd rendering to derive a novel approach. The main limitations we have found when using impostors are (1) the big amount of memory needed to store them, which also has to be sent to the graphics card, (2) the lack of visual quality in close-up views, and (3) some visibility problems. As we wanted to overcome these limitations, and improve performance results, the found conclusions lead us to present a new representation for 3D animated characters using relief mapping, thus supporting an output-sensitive rendering. The basic idea of our approach is to encode each character through a small collection of textured boxes storing color and depth values. At runtime, each box is animated according to the rigid transformation of its associated bone in the animated skeleton. A fragment shader is used to recover the original geometry using an adapted version of relief mapping. Unlike competing output-sensitive approaches, our compact representation is able to recover high-frequency surface details and reproduces view-motion parallax e ects. Furthermore, the proposed approach ensures correct visibility among di erent animated parts, and it does not require us to prede ne the animation sequences nor to select a subset of discrete views. Finally, a user study demonstrates that our approach allows for a large number of simulated agents with negligible visual artifacts.