# MixLight: Borrowing the Best of both Spherical Harmonics and Gaussian Models

**Authors:** Xinlong Ji, Fangneng Zhan, Shijian Lu, Shi-Sheng Huang, Hua Huang  
**arXiv:** [2404.12768](https://arxiv.org/abs/2404.12768)  
**Date:** 2024-04-19  
**Summary page:** https://www.summarizepaper.com/en/paper/2404.12768v1/

## Key points

- Accurate estimation of scene lighting is crucial for creating immersive and realistic mixed reality experiences
- Traditional methods for estimating illumination have limitations in generalization performance and capturing the full spectrum of lighting components
- MixLight is a novel approach that combines Spherical Harmonic (SH) and Spherical Gaussian (SG) models to provide a comprehensive representation of illumination
- MixLight utilizes SH for low-frequency ambient lighting and SG for high-frequency light sources, offering a balanced depiction of scene lighting
- The introduction of the specialized spherical light source sparsemax (SLSparsemax) module enhances the sparsity of spherical light sources by considering their position and brightness relationship
- Extensive experiments demonstrate MixLight's superiority over state-of-the-art methods across multiple metrics, including generalization performance on Web Dataset
- MixLight, developed by authors Xinlong Ji, Fangneng Zhan, Shijian Lu, Shi-Sheng Huang, and Hua Huang, represents a groundbreaking advancement in scene lighting estimation for mixed reality applications

## Summary

, , , , 

In the realm of mixed reality applications, accurately estimating scene lighting is crucial for creating immersive and realistic experiences. Traditional methods for estimating illumination have involved generating illumination maps or regressing illumination parameters. However, these approaches have shown limitations in terms of generalization performance and capturing the full spectrum of lighting components. To address these challenges, a novel approach called MixLight has been introduced in this paper. MixLight is a joint model that leverages the strengths of both Spherical Harmonic (SH) and Spherical Gaussian (SG) models to provide a more comprehensive representation of illumination. By utilizing SH to capture low-frequency ambient lighting and SG to represent high-frequency light sources, MixLight offers a balanced and detailed depiction of scene lighting. Moreover, the paper introduces a specialized spherical light source sparsemax (SLSparsemax) module that enhances the sparsity of spherical light sources by considering their position and brightness relationship. This innovative module fills a significant gap in prior works by improving the efficiency and effectiveness of representing spherical light sources. Extensive experiments conducted to evaluate MixLight demonstrate its superiority over state-of-the-art methods across multiple metrics. Furthermore, experiments on Web Dataset highlight MixLight's superior generalization performance compared to non-parametric methods. The collaborative efforts of authors Xinlong Ji, Fangneng Zhan, Shijian Lu, Shi-Sheng Huang, and Hua Huang have resulted in MixLight: Borrowing the Best of both Spherical Harmonics and Gaussian Models as a groundbreaking advancement in scene lighting estimation for mixed reality applications.