Article
Computer Science, Software Engineering
Jiahui Fan, Beibei Wang, Wenshi Wu, Milos Hasan, Jian Yang, Ling-Qi Yan
Summary: Rendering glinty details from specular microstructure enhances realism in computer graphics. This article proposes a differentiable regularization method to render specular glints, which includes two steps: using differentiable path tracing to render a scene and recording gradients, and predicting the target value by extrapolating the results. The method reduces noise significantly and achieves results close to the reference.
IEEE TRANSACTIONS ON VISUALIZATION AND COMPUTER GRAPHICS
(2023)
Article
Computer Science, Software Engineering
Haowen Tan, Junqiu Zhu, Yanning Xu, Xiangxu Meng, Lu Wang, Ling-Qi Yan
Summary: This method proposes an example-based real-time rendering method to reduce the storage space of high-resolution normal maps and can render complex scenes in real-time.
COMPUTER GRAPHICS FORUM
(2022)
Article
Computer Science, Software Engineering
Jia-Wun Jhang, Chun-Fa Chang
Summary: This paper proposes an improved version of Specular Manifold Sampling (SMS) called Specular Manifold Bisection Sampling (SMBS), which addresses the issues of the original method when the ray deviates too much from the light or bounces from a complex surface. Experimental results demonstrate that SMBS can find more valid manifold paths in fewer iterations, achieving significant improvement in scenes with complex surfaces.
COMPUTER GRAPHICS FORUM
(2022)
Article
Geochemistry & Geophysics
Tamas Varnai, Alexander Marshak, Alexander B. Kostinski
Summary: This letter provides a wider view on the earlier analyses of observed glints caused by clouds, focusing on how the appearance of these glints varies with wavelength and season. The statistical analysis reveals that the wavelength dependence of glints is mainly shaped by the air above the cloud top, and that the radiative impact of cloud glints displays seasonal variations.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Computer Science, Software Engineering
Rui Yu, Yue Dong, Youkang Kong, Xin Tong
Summary: This method introduces a learning-based approach for efficient rendering of pure specular light transport by training a neural network to model the distribution of all specular light paths between pairs of endpoints, resulting in high-quality results and improved rendering speed.
COMPUTER GRAPHICS FORUM
(2023)
Article
Mathematics, Applied
E. Gomez-Deniz, J. M. Sarabia, E. Calderin-Ojeda
Summary: This paper introduces a new family of continuous random variables with non-necessarily symmetric densities, which can capture unimodality and bimodality features, including the normal distribution as a special case. The closed-form density function allows for easy computation of probabilities, moments, skewness, and kurtosis coefficients. Additionally, a stochastic representation of the family is presented for generating random variates. Applications of this new distribution family include explaining the incidence of Hodgkin's disease by age and studying bimodality in geoscience. The multivariate counterpart of this distribution is briefly discussed.
JOURNAL OF COMPUTATIONAL AND APPLIED MATHEMATICS
(2021)
Article
Chemistry, Multidisciplinary
Myungjin Choi, Jee-Hyeok Park, Qimeng Zhang, Byeung-Sun Hong, Chang-Hun Kim
Summary: This study introduced a novel method for generating highly refined normal maps for screen-space fluid rendering using a conditional generative adversarial network. Experimental results showed that the generated features were clearer and more detailed compared to basic methods.
APPLIED SCIENCES-BASEL
(2021)
Article
Computer Science, Software Engineering
Hong Deng, Yang Liu, Beibei Wang, Jian Yang, Lei Ma, Nicolas Holzschuch, Ling-Qi Yan
Summary: This article presents an efficient method for rendering glinty appearance, achieving constant storage and performance through precomputation and data compression. It is applicable to various practical rendering applications.
ACM TRANSACTIONS ON GRAPHICS
(2022)
Article
Engineering, Electrical & Electronic
Yanlong Cao, Xiaoyao Wei, Wenyuan Liu, Binjie Ding, Jiangxin Yang, Yanpeng Cao
Summary: This paper addresses the challenging task of high-quality reconstruction of polished surfaces. The proposed learning-based method achieves a better balance between preserving texture and removing specular outliers compared to state-of-the-art methods. A feature fusion convolutional neural network is used to estimate the surface normal, and a flexible hardware setup is designed for photometric stereo information collection. Experimental results demonstrate that the proposed method can reconstruct the normal of polished surfaces at the pixel level with abundant texture information.
Article
Multidisciplinary Sciences
Sungjin Kim, Chang Ha Lee
Summary: This paper presents a mesh clustering and reordering method based on normal coherence for efficient back-face culling, which has been shown to be more efficient than traditional methods, especially for large and static models. The method first vertically clusters the mesh into multiple stripes based on the latitude of the face normal vector in pre-computation, and computes a potentially visible set of faces at runtime by excluding back faces from the clustered and reordered faces list to improve rendering speed.
Article
Optics
Xu Qian, Jian Yang, Shuo Shi, Wei Gong, Lin Du, Biwu Chen, Bowen Chen
Summary: This study analyzed and corrected the incident angle effect in the whole band of HSL using the Lambert-Beckman model. The results showed that the model efficiently fits the changing of echo intensity with incidence angle and eliminates the specular effect of the target. The Lambert-Beckman model significantly improves the coefficient of variation ratio compared to the reference target-based model, providing advanced insight into correcting the echo intensity in HSL.
Article
Materials Science, Coatings & Films
Dongfei Wang, Jiaqiang Dang, Yugang Li, Zhongming Liu, Haowei Wang, Ming Chen
Summary: The surface integrity of 300M steel was compared after shot peening, laser shock peening, ultrasonic surface rolling process, and hybrid laser shock peening with ultrasonic surface rolling process. The results showed that ultrasonic surface rolling process and hybrid laser shock peening with ultrasonic surface rolling process resulted in a smooth surface, while shot peening and laser shock peening increased surface roughness. The treated specimens exhibited severe plastic deformation and a gradient microstructure with compressive residual stress. Among the different treatments, hybrid laser shock peening with ultrasonic surface rolling process was found to be the most effective surface modification technique.
SURFACE & COATINGS TECHNOLOGY
(2022)
Article
Chemistry, Analytical
Meishu Wang, Su Qiu, Weiqi Jin, Jie Yang
Summary: This study presents a method called De-Glints to suppress water surface glints and obtain clear underwater images using a division of focal plane (DoFP) polarimeter. By calculating the best polarization angle and the image corresponding to the minimal average gray level of each pixel based on polarization imaging principle, the image quality is significantly improved. The experiments demonstrate that the method has an angle calculation error within 10% and can relatively improve the E index by 8.00 under strong glint interference, showing good adaptability to dynamic scenes.
Article
Computer Science, Software Engineering
Artem Komarichev, Jing Hua, Zichun Zhong
Summary: In this paper, a new self-supervised learning method called DiffSVR is proposed, which represents a complicated surface as a depth-aware occupancy function and optimizes it through differentiable surface rendering. Experimental results on benchmark datasets show that DiffSVR outperforms state-of-the-art methods in terms of both numerical performance and visual quality of reconstructed surfaces.
COMPUTER-AIDED DESIGN
(2023)
Article
Physics, Multidisciplinary
Francesco Parisen Toldin, Max A. Metlitski
Summary: It has been recently discovered that the three-dimensional O(N) model exhibits an extraordinary boundary universality class for a finite range of N >= 2. The existence and universal properties of this class for a given N are predicted to be controlled by certain amplitudes of the normal universality class, where a symmetry-breaking field is applied to the boundary. In this study, we investigate the normal universality class for N = 2, 3 through Monte Carlo simulations on an improved lattice model and extract these universal amplitudes. Our results agree well with direct Monte Carlo studies of the extraordinary universality class, providing a nontrivial quantitative confirmation of the relationship between the normal and extraordinary classes.
PHYSICAL REVIEW LETTERS
(2022)