HDGS: Textured 2D Gaussian Splatting for Enhanced Scene Rendering

Yunzhou Song1*, Heguang Lin1*, Jiahui Lei1#, Lingjie Liu1, Kostas Daniilidis1,2
1University of Pennsylvania
2Archimedes, Athena RC
*Equal Contribution, #Corresponding Author
Paper arXiv Code

We illustrate our method's anti-aliasing effect and fine-detail capturing performance compared with 2D Gaussian Splatting. Our method mitigates high-frequency rendering aliases and can capture fine details that other baseline methods fail to preserve.

Abstract

Recent advancements in neural rendering, particularly 2D Gaussian Splatting (2DGS), have shown promising results for jointly reconstructing fine appearance and geometry by leveraging 2D Gaussian surfels. However, current methods face significant challenges when rendering at arbitrary viewpoints, such as anti-aliasing for down-sampled rendering, and texture detail preservation for high-resolution rendering. We proposed a novel method to align the 2D surfels with texture maps and augment it with per-ray depth sorting and fisher-based pruning for rendering consistency and efficiency. With correct order, per-surfel texture maps significantly improve the capabilities to capture fine details. Additionally, to render high-fidelity details in varying viewpoints, we designed a frustum-based sampling method to mitigate the aliasing artifacts. Experimental results on benchmarks and our custom texture-rich dataset demonstrate that our method surpasses existing techniques, particularly in detail preservation and anti-aliasing.

5 min Intro Video with Narration



Method Overview

1. We proposed a coarse-to-fine pipeline to capture fine-detailed textures efficiently. 2. We designed a frustum-based sampling strategy for anti-aliasing rendering. 3. We augmented the per-surfel texture maps with per-ray depth sorting to mitigate popping artifacts.

BibTeX


        @misc{song2024hdgstextured2dgaussian,
      title={HDGS: Textured 2D Gaussian Splatting for Enhanced Scene Rendering}, 
      author={Yunzhou Song and Heguang Lin and Jiahui Lei and Lingjie Liu and Kostas Daniilidis},
      year={2024},
      eprint={2412.01823},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2412.01823}, 
      }