Article
Computer Science, Software Engineering
Jie Guo, Zijing Zong, Yadong Song, Xihao Fu, Chengzhi Tao, Yanwen Guo, Ling-Qi Yan
Summary: This paper proposes a new method for probe-based global illumination (GI) rendering that can generate a wide range of GI effects, including glossy reflection with multiple bounces, in complex scenes. The method utilizes a gradient-based search algorithm and a neural image reconstruction method to achieve fast and high-quality GI rendering, while employing temporal reprojection and a temporal loss to improve temporal stability for animation sequences.
ACM TRANSACTIONS ON GRAPHICS
(2022)
Article
Computer Science, Software Engineering
Zander Majercik, Thomas Muller, Alexander Keller, Derek Nowrouzezahrai, Morgan McGuire
Summary: This study enhances sample efficiency for transport contributions terminating on diffuse scattering events by combining screen-space reservoir resampling and sparse world-space probes, resulting in clear improvements over purely path tracing and probe-based baselines in rendering global illumination interactively in complex scenes.
COMPUTER GRAPHICS FORUM
(2022)
Article
Computer Science, Artificial Intelligence
Hongzhi Jiang, Yuxi Li, Huijie Zhao, Xudong Li, Xu Yang
Summary: Parallel single-pixel imaging (PSI) is a photography technique that captures light transport coefficients and enables 3D shape reconstruction under strong global illumination by extending single-pixel imaging (SI) to modern digital cameras. The local region extension (LRE) method is introduced to accelerate data acquisition of PSI, which is extremely beneficial for data acquisition efficiency. PSI possesses several properties and advantages, such as capturing complete light transport coefficients and not requiring special hardware or restrictions on the arrangement of projector-camera pairs.
INTERNATIONAL JOURNAL OF COMPUTER VISION
(2021)
Article
Computer Science, Software Engineering
Fuyan Liu, Junwen Gan
Summary: In this paper, a novel approach for sampling indirect illumination using bidirectional path tracing and world space resampling reservoir is proposed. Light subpaths with high contribution are cached in a world space grid, allowing for sampling at any vertices along the eye path. Optimal samples are found by continuously updating the reservoirs stored in each cell based on resampling principles. Experimental results show that the algorithm achieves noticeable noise reduction in scenes lit by indirect light compared to path tracing, while maintaining interactive frame rates, albeit with additional time and memory consumption. In some special cases, the method even outperforms a plain bidirectional path tracer in terms of both image quality and performance.
COMPUTERS & GRAPHICS-UK
(2023)
Article
Computer Science, Software Engineering
Tizian Zeltner, Sebastien Speierer, Iliyan Georgiev, Wenzel Jakob
Summary: Physically based differentiable rendering algorithms have advanced in computing derivatives with respect to millions of parameters, but their fundamental properties are still poorly understood and further research is needed. Current algorithms mechanically differentiate a given primal algorithm, limiting potential improvements.
ACM TRANSACTIONS ON GRAPHICS
(2021)
Article
Computer Science, Software Engineering
Shlomi Steinberg, Ling-Qi Yan
Summary: This paper presents the first global light transport framework that takes into account the statistical properties of light and its global evolution, which is fully consistent with Maxwell's electromagnetic theory and retains some attractive properties of traditional radiometry-based light transport.
ACM TRANSACTIONS ON GRAPHICS
(2021)
Article
Computer Science, Information Systems
Wonjun Lee, Piljoong Jeong, Hajin Choi, Jinwoo Kim, Bochang Moon
Summary: The research proposes a technique for estimating scene illumination through online learning to achieve visual consistency between virtual and real objects, capable of providing high-quality global illumination results in augmented reality.
Article
Computer Science, Software Engineering
Duan Gao, Haoyuan Mu, Kun Xu
Summary: We propose a novel method called neural global illumination for fast rendering of full global illumination in static scenes with dynamic viewpoint and area lighting. Our method utilizes a deep rendering network to model the mapping from each shading point to global illumination. We also introduce a neural-network-friendly input representation and a screen-space neural buffer to share global information between objects. This method can synthesize complex global illumination effects efficiently and is capable of handling scenes with multiple-bounce glossy interreflection, color bleeding, and caustics.
IEEE TRANSACTIONS ON VISUALIZATION AND COMPUTER GRAPHICS
(2023)
Article
Computer Science, Software Engineering
Stavros Diolatzis, Julien Philip, George Drettakis
Summary: Neural rendering algorithms with active exploration method using Markov Chain Monte Carlo enable efficient network training and interactive rendering of complex scenes, providing dynamic scene navigation and manipulation.
ACM TRANSACTIONS ON GRAPHICS
(2022)
Article
Computer Science, Software Engineering
Delio Vicini, Sebastien Speierer, Wenzel Jakob
Summary: This paper proposes a new unbiased backpropagation algorithm for rendering that requires constant memory and has computation time linear in the number of scattering events. The approach is based on the invertibility of the local Jacobian at scattering interactions and extends to specular materials that prior work cannot handle.
ACM TRANSACTIONS ON GRAPHICS
(2021)
Article
Computer Science, Software Engineering
Doga Yilmaz, Furkan Kirac
Summary: The realm of 3D computer vision and graphics has grown rapidly, allowing for the creation of realistic virtual environments and digital representations of real-world objects. Current methods for 3D reconstruction predominantly use neural scene representations, but there are limitations such as the need for a large number of captured images and the complexity of transforming these representations into explicit geometric forms. An alternative approach gaining traction is the use of physically-based differentiable rendering (PBDR) and inverse rendering, which require fewer view-points and yield explicit results.
COMPUTERS & GRAPHICS-UK
(2023)
Article
Computer Science, Software Engineering
Arnau Colom, Ricardo Marques, Luis Paulo Santos
Summary: This study proposes a VPL-based ray tracing approach that achieves interactive rendering by clustering and computing visibility on the GPU.
COMPUTERS & GRAPHICS-UK
(2022)
Article
Computer Science, Software Engineering
Jinkai Hu, Milo K. Yip, Guillermo Elias Alonso, Shihao Gu, Xiangjun Tang, Xiaogang Jin
Summary: SDFDDGI is a novel approach for real-time dynamic diffuse global illumination, utilizing signed distance fields (SDF) to approximate scenes with simple primitive shapes and reconstructing irradiance functions in space domain via probes. It supports real-time updates, interpolation, and screen-space refinement for enhanced visual quality. The method efficiently handles dynamic geometry and lighting without precomputation, while also supporting multi-bounced light and managing both large open scenes and indoor high-detailed scenes without hardware restrictions.
Article
Engineering, Electrical & Electronic
Zhizhen Zhao, Jong Chul Ye, Yoram Bresler
Summary: Physics-informed generative modeling is a rapidly growing field in computational imaging, with various methods and applications. This review focuses on generative modeling techniques, such as variational autoencoders (VAEs) and generative adversarial networks (GANs), and recent advancements in score-based generative models. Through different imaging applications, the review demonstrates how these generative modeling techniques effectively incorporate the physics of the imaging problem to solve inverse problems.
IEEE SIGNAL PROCESSING MAGAZINE
(2023)
Article
Optics
Chen Fan, Hong Zhao, Zixin Zhao, Junxiang LI, Yijun DU, Xingyu Yang, Lu Zhang
Summary: We propose a novel single-shot quantitative phase imaging technique using a liquid crystal spatial light modulator (LC-SLM) under white light illumination. By studying the phase modulation characteristics of an LC-SLM under white light illumination, images captured at different wavelengths are equivalent to those captured at different defocus distances. The phase is retrieved from a single-shot color image using the transport of intensity equation. The proposed method shows flexibility and accuracy in quantitative phase imaging of various samples.