Specular Polynomials

Zhimin Fan,Jie Guo,Yiming Wang,Tianyu Xiao,Hao Zhang,Chenxi Zhou,Zhenyu Chen,Pengpei Hong,Yanwen Guo,Ling-Qi Yan

2024-05-22

Abstract:Finding valid light paths that involve specular vertices in Monte Carlo rendering requires solving many non-linear, transcendental equations in high-dimensional space. Existing approaches heavily rely on Newton iterations in path space, which are limited to obtaining at most a single solution each time and easily diverge when initialized with improper seeds.

Graphics

What problem does this paper attempt to address?

This paper attempts to solve the problems encountered when dealing with light paths containing specular chains (i.e., multiple consecutive specular scattering events) in the Monte Carlo rendering algorithm. Specifically, the main bottleneck faced by existing methods in handling these paths is that the light paths of specular reflection / refraction vertices that satisfy all physical constraints have an extremely small probability of being sampled. To solve this problem, previous methods relied on the multivariate Newton - Raphson iteration method, but this method can only obtain one solution at a time and is prone to divergence due to improper selection of the initial seed path, thereby introducing significant bias or variance.

### Core Contributions of the Paper

The paper proposes a new method called "specular polynomial" to solve the above problems in the following ways:

1. **Polynomial Representation of Specular Constraints**:
The authors reformulate the specular constraints as a polynomial system, which simplifies the task into a univariate root - finding problem. Through rational coordinate mapping, the high - dimensional problem is transformed into a low - dimensional problem.

2. **Specular Path Solver without Newton - Raphson Iteration**:
Using the hidden variable resultant method combined with a direct or eigenvalue solver, a deterministic specular path solver that does not rely on multivariate Newton - Raphson iteration is proposed. This method can directly find all eligible specular paths.

3. **Application to the Rendering of Glints and Caustics**:
This method is applied to the rendering of glints and caustics, achieving a fast and almost noise - free simulation of specular light - transport effects. Compared with existing Newton - Raphson - based methods, the new method can handle complex scenes more effectively and reduce artifacts such as strong outliers and energy loss.

### Key Problems Solved

- **Avoiding the Limitations of Newton - Raphson Iteration**: By completely eliminating the dependence on Newton - Raphson iteration, the problems of easy divergence and only being able to find a single solution in the Newton - Raphson method are fundamentally solved.
- **Efficiently Handling Complex Scenes**: The new method can naturally handle discontinuities caused by small specular geometries and complex visibility, reducing artifacts in rendering.
- **GPU - Friendly**: This method is designed to be friendly to CPU and GPU implementations and is suitable for large - scale parallel computing.

### Summary

The main contribution of the paper is to provide a brand - new framework for calculating all eligible specular paths without relying on multivariate Newton - Raphson iteration. This framework combines analytical derivation and numerical calculation, can robustly handle the complexity of specular constraints, and performs excellently in practical applications.