18 results
Search Results
Now showing 1 - 10 of 18
Item Dense 3D Gaussian Splatting Initialization for Sparse Image Data(The Eurographics Association, 2024) Seibt, Simon; Chang, Thomas Vincent Siu-Lung; von Rymon Lipinski, Bartosz ; Latoschik, Marc Erich; Liu, Lingjie; Averkiou, MelinosThis paper presents advancements in novel-view synthesis with 3D Gaussian Splatting (3DGS) using a dense and accurate SfM point cloud initialization approach. We address the challenge of achieving photorealistic renderings from sparse image data, where basic 3DGS training may result in suboptimal convergence, thus leading to visual artifacts. The proposed method enhances precision and density of initially reconstructed point clouds by refining 3D positions and extrapolating additional points, even for difficult image regions, e.g. with repeating patterns and suboptimal visual coverage. Our contributions focus on improving ''Dense Feature Matching for Structure-from-Motion'' (DFM4SfM) based on a homographic decomposition of the image space to support 3DGS training: First, a grid-based feature detection method is introduced for DFM4SfM to ensure a welldistributed 3D Gaussian initialization uniformly over all depth planes. Second, the SfM feature matching is complemented by a geometric plausibility check, priming the homography estimation and thereby improving the initial placement of 3D Gaussians. Experimental results on the NeRF-LLFF dataset demonstrate that this approach achieves superior qualitative and quantitative results, even for fewer views, and the potential for a significantly accelerated 3DGS training with faster convergence.Item Tetrahedral Interpolation on Regular Grids(The Eurographics Association, 2021) Bán, Róbert; Valasek, Gábor; Bittner, Jirí and Waldner, ManuelaThis work proposes the use of barycentric interpolation on enclosing simplices of sample points to infer a reconstructed function from discrete data. In particular, we compare the results of trilinear and tetrahedral interpolation over regular 3D grids of second order spherical harmonics (SH) light probes. In general, tetrahedral interpolation only requires four data samples per query in contrast to the 8 samples necessary for trilinear interpolation, at the expense of a more expensive weight computation. Our tetrahedral implementation subdivides the cubical cells into six tetrahedra and uses the barycentric coordinates of the query position as weights to blend the probe data. We show that barycentric coordinates can be calculated efficiently in shaders for our particular tetrahedral decomposition of the cube, resulting only in simple arithmetic and conditional move operations.Item Stochastic Light Culling for Single Scattering in Participating Media(The Eurographics Association, 2022) Fujieda, Shin; Tokuyoshi, Yusuke; Harada, Takahiro; Pelechano, Nuria; Vanderhaeghe, DavidWe introduce a simple but efficient method to compute single scattering from point and arbitrarily shaped area light sources in participating media. Our method extends the stochastic light culling method to volume rendering by considering the intersection of a ray and spherical bounds of light influence ranges. For primary rays, this allows simple computation of the lighting in participating media without hierarchical data structures such as a light tree. First, we show how to combine equiangular sampling with the proposed light culling method in a simple case of point lights. We then apply it to arbitrarily shaped area lights by considering virtual point lights on the surface of area lights. Using our method, we are able to improve the rendering quality for scenes with many lights without tree construction and traversal.Item Sparse Ferguson-Hermite Signed Distance Fields(The Eurographics Association, 2023) Bán, Róbert; Valasek, Gábor; Singh, Gurprit; Chu, Mengyu (Rachel)We investigate Hermite interpolation in the context of discrete signed distance field filtering. Our method uses tricubic Hermite interpolation to generate a C1 continuous approximation to the signed distance function of the input scene. Our representation is kept purely first order by setting the mixed partial derivatives to zero, similarly to how Ferguson constructed bicubic Hermite patches. Our scheme stores four scalars at each sample, the value of the signed distance function and its first three partial derivatives. We optimize storage by only storing voxels that enclose a volume boundary. We show that this provides both a significant reduction in storage and render times compared to a dense grid of Ferguson-Hermite samples. Moreover, our construct requires smaller storage than traditional zero order trilinearly filtered fields of the same visual quality, at the expense of performance.Item Illumination-driven Light Probe Placement(The Eurographics Association, 2021) Vardis, Konstantinos; Vasilakis, Andreas Alexandros; Papaioannou, Georgios; Bittner, Jirí and Waldner, ManuelaWe introduce a simplification method for light probe configurations that preserves the indirect illumination distribution in scenes with diverse lighting conditions. An iterative graph simplification algorithm discards the probes that, according to a set of evaluation points, have the least impact on the global light field. Our approach is simple, generic and aims at improving the repetitive and often non-intuitive and tedious task of placing light probes on complex virtual environments.Item Radiance-Based Blender Add-On for Physically Accurate Rendering of Cultural Heritage(The Eurographics Association, 2023) Méndez, Míriam; Munoz-Pandiella, Imanol; Andujar, Carlos; Singh, Gurprit; Chu, Mengyu (Rachel)Despite the Cultural Heritage and Computer Graphics communities are increasingly joining forces to strengthen their collaboration, the study of how light interacts with monuments (e.g. weathering the surfaces or affecting the visitors' experience) is still an open problem in cultural heritage. A significant limitation is the lack of easy-to-use, open-source, physically-accurate tools allowing cultural heritage experts to perform lighting simulations on the increasing collection of 3D reconstructions. In this work, we present an open-source Blender add-on to facilitate such simulations. The add-on allows art historians to configure the properties (materials, lights, and camera) of the simulation, and uses as rendering back-end the Radiance software, a validated physically accurate light simulation tool. Our tool lowers the entry barrier for the use of a highly accurate but rather complex (command-based) tool for lighting studies in cultural heritage monuments.Item Simple Techniques for a Novel Human Body Pose Optimisation Using Differentiable Inverse Rendering(The Eurographics Association, 2022) Battogtokh, Munkhtulga; Borgo, Rita; Pelechano, Nuria; Vanderhaeghe, DavidHuman body 3D reconstruction has a wide range of applications including 3D-printing, art, games, and even technical sport analysis. This is a challenging problem due to 2D ambiguity, diversity of human poses, and costs in obtaining multiple views. We propose a novel optimisation scheme that bypasses the prior bias bottleneck and the 2D-pose annotation bottleneck that we identify in single-view reconstruction, and move towards low-resource photo-realistic 3D reconstruction that directly and fully utilises the target image. Our scheme combines domain-specific method SMPLify-X and domain-agnostic inverse rendering method redner, with two simple yet powerful techniques. We demonstrate that our techniques can 1) improve the accuracy of the reconstructed body both qualitatively and quantitatively for challenging inputs, and 2) control optimisation to a selected part only. Our ideas promise extension to more difficult problems and domains even beyond human body reconstruction.Item An Improved Triangle Encoding Scheme for Cached Tessellation(The Eurographics Association, 2022) Kerbl, Bernhard; Horváth, Linus; Cornel, Daniel; Wimmer, Michael; Pelechano, Nuria; Vanderhaeghe, DavidWith the recent advances in real-time rendering that were achieved by embracing software rasterization, the interest in alternative solutions for other fixed-function pipeline stages rises. In this paper, we revisit a recently presented software approach for cached tessellation, which compactly encodes and stores triangles in GPU memory. While the proposed technique is both efficient and versatile, we show that the original encoding is suboptimal and provide an alternative scheme that acts as a drop-in replacement. As shown in our evaluation, the proposed modifications can yield performance gains of 40% and more.Item Visualizing Errors in Rendered High Dynamic Range Images(The Eurographics Association, 2021) Andersson, Pontus; Nilsson, Jim; Shirley, Peter; Akenine-Möller, Tomas; Theisel, Holger and Wimmer, MichaelA new error metric targeting rendered high dynamic range images is presented. Our method computes a composite visualization over a number of low dynamic range error maps of exposure compensated and tone mapped image pairs with automatically computed, or manually provided, parameters. We argue that our new error maps predict errors substantially better than metrics previously used in rendering. Source code is released with the hope that our work can be a useful tool for future research.Item Hermite Interpolation of Heightmaps(The Eurographics Association, 2022) Bán, Róbert; Valasek, Gábor; Sauvage, Basile; Hasic-Telalovic, JasminkaHeightmaps are ubiquitous in real-time computer graphics. They are used to describe geometric detail over an underlying coarser surface. Various techniques, such as parallax occlusion mapping and relief mapping, use heightmap textures to impose mesostructural details over macrostructural elements without increasing the actual complexity of the rendered geometries. We aim to improve the quality of the fine resolution surface by incorporating the gradient of the original function into the sampling procedure. The traditional representation consists of simple height values stored on a regular grid. During rendering, bilinear filtering is applied. We propose to store the partial derivatives with the height values and use Hermite interpolation between the samples. This guarantees a globally C1 continuous heightfield instead of the C0 -continuity of bilinear filtering. Moreover, incorporating higher order information via partial derivatives allows us to use lower resolution heightmaps while retaining the appearance of a higher resolution map. In parallax mapping, surface normals are often stored alongside the height values, as such, our method does not require additional storage, since normals and partial derivatives can be calculated from one another. The exact normals of the reconstructed cubic Hermite surface can also be calculated, resulting in a storage efficient replacement for normal mapping with richer visual appearance.