WishGI: Lightweight Static Global Illumination Baking via Spherical Harmonics Fitting 1332 Journal: TOG Volume: 44 4 8 DOI: 10.1145/3730935 CCS: Computing methodologies Rendering
Junke Zhu, Zehan Wu, Qixing Zhang, Cheng Liao, Zhangjin Huang · 2025
Read WishGI: Lightweight Static Global Illumination Baking via Spherical Harmonics Fitting 1332 Journal: TOG Volume: 44 4 8 DOI: 10.1145/3730935 CCS: Computing methodologies Rendering by Junke Zhu, Zehan Wu, Qixing Zhang, Cheng Liao, Zhangjin Huang (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-01288.txt.
Introduction
Junke Zhu, Zehan Wu, Qixing Zhang, Cheng Liao, and Zhangjin Huang present WishGI, a lightweight method for baking static global illumination in rendered scenes. Global illumination combines direct and indirect light for realistic images, and precomputing it cuts runtime costs. Many existing methods, though, rely on extensive texture storage and heavy sampling, which low-end platforms cannot afford.
WishGI reconstructs illumination in a way that reduces fragment-shader sampling and avoids extra render passes. It bakes illumination information by fitting spherical harmonics, and it proposes an inverse probe distribution that assigns each mesh its own probe associations, generated offline in local space. That keeps lighting quality consistent across every instance of the same mesh. The authors report highly competitive lighting effects while using only about five percent of the memory mainstream industry techniques need. The method targets games and interactive graphics on modest hardware.
Catalog record
| CREATOR | Junke Zhu, Zehan Wu, Qixing Zhang, Cheng Liao, Zhangjin Huang |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 45 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-01288
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
WishGI: Lightweight Static Global Illumination Baking via Spherical Harmonics Fitting 1332 Journal: TOG Volume: 44 4 8 DOI: 10.1145/3730935 CCS: Computing methodologies Rendering Junke Zhu, Zehan Wu, Qixing Zhang, Cheng Liao, Zhangjin Huang Abstract . Global illumination combines direct and indirect lighting to create realistic lighting effects, bringing virtual scenes closer to reality. Static global illumination is a crucial component of virtual scene rendering, leveraging precomputation and baking techniques to significantly reduce runtime computational costs. Unfortunately, many existing works prioritize visual quality by relying on extensive texture storage and massive pixel-level texture sampling, leading to large performance overhead. In this paper, we introduce an illumination reconstruction method that effectively reduces sampling in fragment shader and avoids additional render passes, making it well-suited for low-end platforms. To achieve high-quality global illumination with reduced memory usage, we adopt a spherical harmonics fitting approach for baking effective illumination information and propose an inverse probe distribution method that generates unique probe associations for each mesh. This association, which can be generated offline in the local space, ensures consistent lighting quality across all instances of the same mesh. As a consequence, our method delivers highly competitive lighting effects while using only approximately 5% of the memory required by mainstream industry techniques. Figure 1. We present a novel method for static global illumination baking tailored for low-end platforms. Our work significantly reduces memory usage and ensures high runtime efficiency, while avoiding additional render passes, making it highly compatible with forward rendering. This method enables high-quality static global illumination on low-end devices, demonstrating clear advantages over mainstream industry methods. † † cc-license: by 1. Introduction Global illumination (GI) refers to the process of simulating all light interactions within a scene, including both direct and indirect lighting ( Ritschel et al. 2012 ) . High-quality global illumination techniques can make virtual scenes more realistic, providing users with an enhanced experience. In most video games, static objects make up the majority of the scene. By leveraging precomputation and baking techniques, runtime computational costs can be significantly reduced. Therefore, static global illumination baking techniques ( Diolatzis et al. 2022 ; Li et al. 2023 ; Rainer et al. 2022 ) have attracted a great deal of attention. Among various static global illumination techniques, we focus on methods tailored for low-end platforms. Due to hardware limitations, these platforms often face a great challenge of memory and performance. Existing probe-based works ( Games 2024c ; Guo et al. 2022 ; Song and Funkhouser 2019 ; Michal Iwanicki and Shirley 2024 ) require storing a large number of spherical harmonics in the scene and rely on massive pixel-level interpolation during rendering, posing significant challenges to low-end platforms in terms of storage and runtime performance. Epic Games ( Games 2024a ) just distributes probes on the surface of objects and reduces runtime overhead through per-component interpolation, making it feasible for mobile platforms. However, its lighting quality deteriorates when applied to structurally complex objects. Sloan et al. ( Sloan and Silvennoinen 2018 ) calculate illumination based on the surface of the object and simplify the storage into light intensity and dominant light direction, which becomes the most widely adopted methods in industry. However, it introduces visual imperfections in low-end platforms due to limited precision and compression artifacts. Moreover, the gaps caused by the use of UV mapping result in wasted memory space. In this paper, we introduce a novel illumination reconstruction method, which supports Level of Detail (LOD), effectively reduces sampling in fragment shader and avoids additional render passes. Therefore, this method is well-suited for forward rendering (a rendering technique widely used in Tile-based Deferred Rendering GPUs). To achieve high-quality global illumination with minimal memory usage, we adopt a spherical harmonics fitting approach that efficiently bakes effective illumination information to ensure accuracy. Since existing probe-based works typically place fixed probes in the scene leading to redundancy in spherical harmonics when applied to baking static objects, we propose an inverse probe distribution method that generates unique probe associations for each mesh in local space through optimization. This association, which can be generated offline, ensures consistent lighting quality across all instances of the same mesh. We demonstrate our work through illumination quality and performance statistics. The major contributions of this work are as follows. • We propose an illumination reconstruction method that effectively reduces sampling in fragment shader and avoids additional render passes, ensuring exceptional runtime performance. • We propose an illumination baking method that bakes effective lighting information via spherical harmonics fitting, improving the utilization of effective information and achieving higher precision. • We propose an inverse probe distribution method that generates unique probe associations for each mesh in the local space to reduce spherical harmonics redundancy. This association can be generated offline and ensures consistent lighting quality across all instances of the same mesh. 2. RELATED WORK Hemispherical illumination. The illumination of objects in the real world usually varies with changes in surface normals. To achieve detailed lighting effects, illumination information must be recorded in the hemispherical space defined by the geometric normals. Thus, spherical harmonics (SH) have been adopted as a medium for representing illumination since earlier works ( Good and Taylor 2005 ; Chen and Liu 2008 ) . Sloan et al. 2002 also introduced the SH-based concept of precomputed radiance transfer, and subsequent works ( Sloan et al. 2005 ; Li et al. 2019 ; Rainer et al. 2022 ) achieved remarkable visual results. However, this technology is far from being supported on low-end platforms due to its significant memory requirements. Simpler representations, such as AHD (Ambient/Highlight/Direction) ( Iwanicki and Sloan 2017b ; Lazarov 2011 ; Iwanicki 2013 ) , significantly reduce storage requirements compared to SH, but AHD’s nonlinear nature makes linear interpolation of individual components mathematically inaccurate. Environment cube-based representations ( Iwanicki and Sloan 2017a ; Hooker 2016 ; Gilabert and Stefanov 2012 ) also provide a good way to represent illumination by storing limited surface data and interpolating to compute colors for arbitrary directions, but they still lag behind spherical harmonics in terms of quality. Our work adopts spherical harmonics to encode hemispherical illumination information at different sample points, preserving more lighting details. Their linearity ensures the feasibility and accuracy of surface interpolation, making the entire baking process differentiable and facilitating subsequent optimization. Global illumination baking. 2D lightmap is one of the most commonly used methods for storing precomputed lighting. Research in this area includes parameterization ( Liu et al. 2008 ) , packing and compression ( 1036 ; Knodt et al. 2023 ) , and seam handling ( Liu et al. 2017 ; Ray et al. 2010 ) . Sloan and Silvennoinen 2018 proposed Directional Lightmap, which stores lighting intensity and the direction of the strongest incoming light. While lightmaps strike a good balance between visual quality and runtime performance, visual imperfections are often introduced on low-end platforms due to limited resolution and compression artifacts. Moreover, the gaps caused by the use of UV mapping result in wasted memory usage. Xing et al. 2024 attempted to store incoming radiance information in point clouds, yet this approach still incurred considerable storage demands, posing challenges for complex scenes. The lightweight probe-based method Indirect Light Caching ( Games 2024a ) is