Adaptive Multiple Control Variates for Many-Light Rendering

Xiaofeng Xu, Lu Wang
Shandong University, China

Comparison to Salaün et al. [2022]

Equal-time comparison among our method (fixed and adaptive number of control variates) with polynomials of order 1, conventional Monte Carlo (MC) integration, and regression-based MC integration of Salaün et al. with polynomials of order 1 (O1), 2 (O2) 3 (O3).

Veach-mis

Living-room2

Bathroom

Staircase2

Classroom

Cornell-box

Comparison to kondapaneni et al. [2019]

Equal-spp comparison among our method, conventional Monte Carlo (MC) integration, the balance and power heuristic (Veach [1997] (Robust Monte Carlo Methods for Light Transport Simulation)), and the optimal weights (Kondapaneni et al.[2019] (Optimal Multiple Importance Sampling)).

Veach-mis

Staircase1

Comparison to Estevez et al. [2019]

Equal-sample comparison of adaptive tree splitting for many lights sampling by Estevez et al. [CEK18] and our method.

Bathroom

Staircase2

Classroom

Cornell-box

Antialiasing

As the number of represented points increases, the degree of aliasing decreases and the time cost increases.

Bathroom

Staircase2