HRGS: Hierarchical Gaussian Splatting for Memory-Efficient High-Resolution 3D Reconstruction

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Li, Changbai, Zhu, Haodong, Chen, Hanlin, Zhang, Juan, Chen, Tongfei, Yang, Shuo, Shao, Shuwei, Dong, Wenhao, Zhang, Baochang
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913898323509248
author Li, Changbai
Zhu, Haodong
Chen, Hanlin
Zhang, Juan
Chen, Tongfei
Yang, Shuo
Shao, Shuwei
Dong, Wenhao
Zhang, Baochang
author_facet Li, Changbai
Zhu, Haodong
Chen, Hanlin
Zhang, Juan
Chen, Tongfei
Yang, Shuo
Shao, Shuwei
Dong, Wenhao
Zhang, Baochang
contents 3D Gaussian Splatting (3DGS) has made significant strides in real-time 3D scene reconstruction, but faces memory scalability issues in high-resolution scenarios. To address this, we propose Hierarchical Gaussian Splatting (HRGS), a memory-efficient framework with hierarchical block-level optimization. First, we generate a global, coarse Gaussian representation from low-resolution data. Then, we partition the scene into multiple blocks, refining each block with high-resolution data. The partitioning involves two steps: Gaussian partitioning, where irregular scenes are normalized into a bounded cubic space with a uniform grid for task distribution, and training data partitioning, where only relevant observations are retained for each block. By guiding block refinement with the coarse Gaussian prior, we ensure seamless Gaussian fusion across adjacent blocks. To reduce computational demands, we introduce Importance-Driven Gaussian Pruning (IDGP), which computes importance scores for each Gaussian and removes those with minimal contribution, speeding up convergence and reducing memory usage. Additionally, we incorporate normal priors from a pretrained model to enhance surface reconstruction quality. Our method enables high-quality, high-resolution 3D scene reconstruction even under memory constraints. Extensive experiments on three benchmarks show that HRGS achieves state-of-the-art performance in high-resolution novel view synthesis (NVS) and surface reconstruction tasks.
format Preprint
id arxiv_https___arxiv_org_abs_2506_14229
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle HRGS: Hierarchical Gaussian Splatting for Memory-Efficient High-Resolution 3D Reconstruction
Li, Changbai
Zhu, Haodong
Chen, Hanlin
Zhang, Juan
Chen, Tongfei
Yang, Shuo
Shao, Shuwei
Dong, Wenhao
Zhang, Baochang
Computer Vision and Pattern Recognition
Artificial Intelligence
3D Gaussian Splatting (3DGS) has made significant strides in real-time 3D scene reconstruction, but faces memory scalability issues in high-resolution scenarios. To address this, we propose Hierarchical Gaussian Splatting (HRGS), a memory-efficient framework with hierarchical block-level optimization. First, we generate a global, coarse Gaussian representation from low-resolution data. Then, we partition the scene into multiple blocks, refining each block with high-resolution data. The partitioning involves two steps: Gaussian partitioning, where irregular scenes are normalized into a bounded cubic space with a uniform grid for task distribution, and training data partitioning, where only relevant observations are retained for each block. By guiding block refinement with the coarse Gaussian prior, we ensure seamless Gaussian fusion across adjacent blocks. To reduce computational demands, we introduce Importance-Driven Gaussian Pruning (IDGP), which computes importance scores for each Gaussian and removes those with minimal contribution, speeding up convergence and reducing memory usage. Additionally, we incorporate normal priors from a pretrained model to enhance surface reconstruction quality. Our method enables high-quality, high-resolution 3D scene reconstruction even under memory constraints. Extensive experiments on three benchmarks show that HRGS achieves state-of-the-art performance in high-resolution novel view synthesis (NVS) and surface reconstruction tasks.
title HRGS: Hierarchical Gaussian Splatting for Memory-Efficient High-Resolution 3D Reconstruction
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2506.14229