Advances in 3D Scene Reconstruction and Geometry Processing

The field of 3D scene reconstruction and geometry processing is rapidly advancing, with a focus on developing novel methods for realistic and view-consistent scene representation. Recent research has explored the use of Gaussian Splatting, Neural Radiance Fields, and other techniques to improve the accuracy and efficiency of scene reconstruction. One notable direction is the integration of multiple types of geometrical primitives, such as Gaussian ellipses and ellipsoids, to better represent complex object surfaces. Additionally, the use of wavelet decomposition and other optimization techniques has shown promise in improving the quality of reconstructed scenes. Noteworthy papers in this area include RePaintGS, which proposes a novel 3D scene inpainting method that reliably produces realistic and perceptually consistent results, and Wavelet-GS, which integrates wavelet decomposition into 3D Gaussian Splatting to improve the quality of complex scene reconstruction. Other notable works, such as OffsetCrust and A Mixed-Primitive-based Gaussian Splatting Method, have also made significant contributions to the field of geometry processing.

Sources

RePaintGS: Reference-Guided Gaussian Splatting for Realistic and View-Consistent 3D Scene Inpainting

Minimum-norm interpolation for unknown surface reconstruction

From images to properties: a NeRF-driven framework for granular material parameter inversion

Legendre Polynomials and Their Use for Karhunen-Lo\`eve Expansion

OffsetCrust: Variable-Radius Offset Approximation with Power Diagrams

A Mixed-Primitive-based Gaussian Splatting Method for Surface Reconstruction

Wavelet-GS: 3D Gaussian Splatting with Wavelet Decomposition

A Unified Framework for Efficient Kernel and Polynomial Interpolation

NeuraLeaf: Neural Parametric Leaf Models with Shape and Deformation Disentanglement

Built with on top of