Density Estimation Techniques for Robust Light Transport Simulation

• Main Authors: Tzu-Chieh Chang, 張子捷
• Other Authors: Ming Ouhyoung
• Format: Others
• Language: en_US
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/xgqnge

碩士 === 國立臺灣大學 === 資訊工程學研究所 === 107 === We propose a robust light transport algorithm that is able to capture all possible transport paths where the vast majority of the techniques involved are based on density estimation. It builds on top of existing progressive photon mapping methods where higher frequency lighting effects are added in subsequent iterations. Our approach is capable of eliminating all singularities encountered in a scene representation, and does not require evaluating material properties during the energy estimation process unlike the rest of photon mapping methods. The proposed photon cone focusing scheme is a consistent method that requires only finite memory to converge. A formal proof of convergence is given along with asymptotic analysis of estimation variance and bias. We compare our results to recent consistent and unbiased methods with scenes that contain difficult light paths in both qualitative and quantitative ways.