基于线性变换余弦球面分布的光照估计方法

doi:10.12068/j.issn.1005-3026.2025.20240030

摘要/Abstract

摘要：

为了精确描述和参数化场景光源，实现高精度的单一图像光照估计，提出了一种基于线性变换余弦球面分布的光照表示方法，构建了回归神经网络，用于从单一图像中推断光源的参数化分布和强度.创新性地引入基于奇异值分解的损失函数，该函数可以精确简洁地衡量两个参数化光源的距离，能够显著提升回归网络的精度.实验结果表明，与现有方法相比，该方法在复杂光照条件下表现优异，尤其在捕捉光照各向异性信息方面有明显改进.

关键词: 光照估计, 线性变换, 余弦球面分布, 回归网络, 奇异值分解

Abstract:

To accurately describe and parameterize scene light sources and achieve high-precision single-image illumination estimation， an illumination representation method based on linear transformation of cosine spherical distribution was proposed. A regression neural network was designed to infer the parametric distribution and intensity of light sources from a single image. A loss function based on singular value decomposition was innovatively introduced. This function could precisely and succinctly measure the distance between two parameterized light sources， significantly enhancing the accuracy of the regression network. Experimental results demonstrate that，compared with existing methods， this method performs exceptionally well under complex illumination conditions， particularly showing a notable improvement in capturing anisotropic illumination information.

Key words: illumination estimation, linear transformation, cosine spherical distribution, regression network, singular value decomposition

中图分类号:

TP 391.4

于连江, 刘洪娟. 基于线性变换余弦球面分布的光照估计方法[J]. 东北大学学报（自然科学版）, 2025, 46(9): 34-40.

Lian-jiang YU, Hong-juan LIU. Illumination Estimation Method Based on Linear Transformation of Cosine Spherical Distribution[J]. Journal of Northeastern University(Natural Science), 2025, 46(9): 34-40.

图/表 7

图1 基于线性变换余弦球面分布的LTCLight光照估计方法框架

Fig.1 Framework of LTCLight illumination estimation method based on linear transformation of cosine spherical distribution

图2 线性变换矩阵控制下的多种高频光照分布

Fig.2 Various high-frequency illumination distributions controlled by linear transformation matrix

图3 矩阵M线性变换的SVD分解的旋转与缩放示意图

Fig.3 Schematic diagram of SVD decomposition, rotation and scaling of linear transformation represented by matrix M

表1 LTC光照表示法与SG和SH方法的误差比较

Table 1 Comparison of errors between LTC illumination representation, SG and SH methods

评估指标	SG(196个锚点)			SH(10阶)			LTC(10个光源)
评估指标	DM	SM	MM	DM	SM	MM	DM	SM	MM
RSME	6.125	6.527	6.752	7.087	7.248	7.301	5.789	5.871	6.097
si-RSME	0.464	0.445	1.219	0.310	0.360	0.696	0.207	0.258	0.519
AE/（°）	12.26	12.83	25.12	12.79	14.72	20.94	9.01	9.65	15.14

图4 真实光照与LTC, SG, SH方法生成光源对比图

Fig.4 Comparison of real illumination and light sources generated by LTC, SG and SH methods

表2 LTCLight与EMLight的回归网络和Gardner 2019方法的误差比较

Table 2 Comparison of errors between LTCLight, EMLight regression network and Gardner 2019 method

评估指标	EMLight的回归网络			Gardner 2019			LTCLight
评估指标	DM	SM	MM	DM	SM	MM	DM	SM	MM
RSME	7.089	7.049	6.970	7.155	7.147	6.822	6.971	6.841	6.528
si-RSME	0.373	0.353	1.019	0.424	0.456	1.257	0.309	0.310	0.615
AE/（°）	15.32	14.26	25.78	16.62	16.91	31.48	13.06	12.98	18.91

图5 真实光照与EMLight的回归网络,Gardner 2019和LTCLight方法回归参数化光源对比图

Fig.5 Comparison of regression parameterized light sources between real illumination and EMLight regression network, Gardner 2019 and LTCLight method

参考文献 19

[1]	Debevec P E. Rendering synthetic objects into real scenes： bridging traditional and image-based graphics with global illumination and high dynamic range photography［C］// Special Interest Group on Graphics and Interactive Techniques. New York： Association for Computing Machinery， 1998： 189-198.
[2]	Karsch K， Hedau V， Forsyth D， et al. Rendering synthetic objects into legacy photographs［J］. ACM Transactions on Graphics， 2011， 30（6）： 1-12.
[3]	Marschner S R， Donald P G. Inverse lighting for photography ［C］//Color and Imaging Conference. Scottsdale. Arizona： International Society for Optical Engineering， 1997： 262-265.
[4]	Tocci M D， Kiser C， Tocci N， et al. A versatile HDR video production system［J］. ACM Transactions on Graphics， 2011， 30（4）： 1-10.
[5]	Barron J T， Malik J. Intrinsic scene properties from a single RGB-D image［J］. IEEE Transactions on Pattern Analysis and Machine Intelligence， 2016， 38（4）： 690-703.
[6]	Wu C L， Wilburn B， Matsushita Y， et al. High-quality shape from multi-view stereo and shading under general illumination［C］//CVPR 2011. Colorado： Springs， 2011： 969-976.
[7]	Liu B， Xu K， Martin R R. Static scene illumination estimation from videos with applications［J］. Journal of Computer Science and Technology， 2017， 32（3）： 430-442.
[8]	Lombardi S， Nishino K. Reflectance and illumination recovery in the wild［J］. IEEE Transactions on Pattern Analysis and Machine Intelligence， 2016， 38（1）： 129-141.
[9]	Maier R， Kim K， Cremers D， et al. Intrinsic3D： high-quality 3D reconstruction by joint appearance and geometry optimization with spatially-varying lighting［C］//2017 IEEE International Conference on Computer Vision （ICCV）. Venice， 2017： 3133-3141.
[10]	Zhan F N， Yu Y C， Zhang C G， et al. GMLight： lighting estimation via geometric distribution approximation［J］. IEEE Transactions on Image Processing， 2022， 31： 2268-2278.
[11]	Wang G C， Yang Y N， Loy C C， et al. StyleLight： HDR panorama generation for lighting estimation and editing［C］//Computer Vision-ECCV 2022. Cham： Springer Nature Switzerland， 2022： 477-492.
[12]	Karimi D M R， Eisenmann J， Hold-Geoffroy Y， et al. EverLight： indoor-outdoor editable HDR lighting estimation［C］//2023 IEEE/CVF International Conference on Computer Vision （ICCV）. Paris， 2023： 7386-7395.
[13]	Garon M， Sunkavalli K， Hadap S， et al. Fast spatially-varying indoor lighting estimation［C］//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition （CVPR）. Long Beach， 2019： 6901-6910.
[14]	Gardner M A， Hold-Geoffroy Y， Sunkavalli K， et al. Deep parametric indoor lighting estimation［C］//2019 IEEE/CVF International Conference on Computer Vision （ICCV）. Seoul， 2019： 7175-7183.
[15]	Li M T， Guo J， Cui X F， et al. Deep spherical Gaussian illumination estimation for indoor scene［C］//Proceedings of the ACM Multimedia Asia. Beijing， 2019： 1-6.
[16]	Heitz E， Dupuy J， Hill S， et al. Real-time polygonal-light shading with linearly transformed cosines［J］. ACM Transactions on Graphics， 2016， 35（4）： 1-8.
[17]	Gardner M A， Sunkavalli K， Yumer E， et al. Learning to predict indoor illumination from a single image［J］. ACM Transactions on Graphics， 2017， 36（6）： 1-14.
[18]	LeGendre C， Ma W C， Fyffe G， et al. DeepLight： learning illumination for unconstrained mobile mixed reality［C］//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition （CVPR）. Long Beach， 2019： 5911-5921.
[19]	Zhan F N， Zhang C G， Yu Y C， et al. EMLight： lighting estimation via spherical distribution approximation［C］// Proceedings of the AAAI Conference on Artificial Intelligence. California： AAAI， 2021： 3287-3295.