3D Gaussian Splatting (3DGS) enables real-time novel view synthesis, but Gaussian rasterization attenuates fine-scale scene structure through the projected covariance of each primitive. We show that contracting these projected covariances at inference time can mitigate the effective bandwidth limitation induced by finite Gaussian footprints, motivating a multi-aperture rendering framework for super-resolution. Based on this insight, we propose AGSR (Aperture-Guided Super-Resolution), a lightweight inference-time framework that progressively fuses multi-aperture renderings using uncertainty-guided feature aggregation and geometric conditioning from transmittance and depth cues. AGSR operates as a plug-in enhancement for pre-trained 3DGS models. Experiments on the Mip-NeRF 360 benchmark show that AGSR outperforms lightweight image-space super-resolution baselines, with gains increasing as the underlying Gaussian representation deteriorates. The results provide evidence that covariance contraction exposes bandwidth-sensitive structural cues attenuated by native Gaussian rasterization. AGSR achieves these improvements while maintaining real-time performance with up to 139k parameters.
Figures & tables
Fig. 1 : Projected Gaussian covariances influence the effective bandwidth of individual rasterization kernels. By systematically contracting the covariance, AGSR generates multi-aperture observations that reveal structural cues attenuated by native rasterization. The resulting uncertainty field guides geometry-aware super-resolution for enhanced detail reconstruction.
Fig. 2 : Overview of AGSR. Covariance contraction produces multi-aperture RGB, transmittance, and depth renderings with progressively smaller projected footprints and correspondingly higher effective bandwidth. Cross-aperture variations form a structural uncertainty field that guides a three-stage geometry-aware fusion network. Higher-bandwidth observations are fused sequentially and refined through residual PixelShuffle reconstruction to produce the final super-resolved output.
Fig. 3 : Visual comparison of 2× and 4× super-resolution on 3DGS representations trained on highly subsampled images ( tf=8 ). AGSR exploits structural cues revealed by multi-aperture rendering to recover fine-scale detail that is poorly reproduced by degraded Gaussian representations. Quantitative metrics (PSNR / SSIM / LPIPS) are reported below each crop.
AGSR 2× ( 960×540→ FHD)
AGSR 4× ( 480×270→ FHD)
Active Stages
Params (K)
GMACs
FPS (FP32)
FPS (FP16)
Params (K)
GMACs
FPS (FP32)
FPS (FP16)
1
21
10.6
40
43
36
4.7
103
112
2
52
26.9
31
35
83
10.8
84
94
3
92
47.6
22
27
139
18.0
64
76
TABLE I : Complexity analysis of the progressive aperture fusion network. We report the parameter count, computational complexity (GMACs), and full-pipeline inference speed (FPS) in both FP32 and FP16 precisions. Parameter counts and GMACs correspond to the cumulative deployed inference model required to produce each stage’s output. FPS is evaluated targeting Full HD ( 1920×1080 ) and includes both 3DGS rasterization and network inference overhead.
tf=2
tf=4
tf=8
Method
PSNR ↑
SSIM ↑
LPIPS ↓
PSNR ↑
SSIM ↑
LPIPS ↓
PSNR ↑
SSIM ↑
LPIPS ↓
Native 3DGS
28.74
0.857
0.260
25.83
0.799
0.299
21.15
0.631
0.437
Target Resolution: 2× Super-Resolution
Bicubic 2×
26.43
0.810
0.294
28.18
0.839
0.270
23.80
0.725
0.358
SPAN-finetuned 2×
24.40
0.742
0.352
26.59
0.819
0.275
21.07
0.628
0.426
SAFMN-finetuned 2×
27.16
0.830
0.281
26.62
0.820
0.268
22.27
0.674
0.387
TABLE II : Quantitative comparison across multiple training degradation factors ( tf ). Best performance highlighted in bold .
Fig. 4 : Quantitative performance (PSNR, SSIM, and LPIPS) of AGSR compared with native 3DGS and bicubic baselines across varying training degradation factors ( tf ). As the training factor increases, the quality of the underlying Gaussian representation deteriorates substantially. AGSR becomes increasingly advantageous as the underlying representation degrades, with the strongest gains observed under moderate and severe degradation, outperforming the baselines across most settings and metrics.
3D Gaussian Splatting (3DGS) enables real-time novel view synthesis, but existing general-purpose acceleration methods suffer severe rendering quality degradation when extended to more complex, large-scale scenes. To address this issue, we propose EffGS, a more general acceleration framework that improves training and rendering efficiency while maintaining reconstruction quality comparable to or better than vanilla 3DGS across bounded and large-scale scenes. EffGS combines frequency-aware guidance, localized density control, and adaptive primitive scale modulation. First, an importance scoring mechanism combines pixel-wise reconstruction errors with a difference-of-Gaussians mask scheduled over training to provide stage-dependent spatial guidance. Second, localized densification and pruning restricts density modifications to Gaussians with valid projected footprints in the sampled views. Third, learnable per-Gaussian scale modulation adjusts effective primitive extent during optimization while retaining the Compact Box rasterization rule. Extensive experiments on bounded and large-scale scene datasets demonstrate a favorable balance between reconstruction quality, training time, and primitive count. Component ablations and matched-primitive-budget comparisons further support the effectiveness of the framework.
Changbai Li, Shuo Yang, Yichen Yang +2
Beihang University · Nanyang Technological University
3D Gaussian Splatting (3DGS) has become the standard for real-time novel view synthesis on commodity GPUs. Its pipeline ties spatial partitioning and rasterization to one tile size, yet the two pull in opposite directions: partitioning, which bins and depth-sorts gaussians, grows cheaper with larger tiles, while rasterization gets cheaper with smaller ones. Prior acceleration work reduces the cost of individual stages but keeps both locked to that single scale, where a few dense tiles dominate frame time. We present Hierarchically Tiled Gaussian Splatting (HiGS), which gives each its own scale: partitioning runs over coarse macro-tiles, while rasterization runs over the fine render tiles within them. Rasterization work is then issued in proportion to the gaussians in each macro-tile rather than per tile, so dense regions spread across many parallel units instead of serializing through one. Across tested scenes, HiGS renders up to 15.8x faster than the original 3DGS and outperforms every other rasterizer we evaluate, while preserving exact front-to-back alpha compositing.
3D Gaussian Splatting (3DGS) enables high-quality real-time 3D rendering but faces challenges in efficiently scaling to ultra-dense scenes and high-resolution due to computational bottlenecks that limit its use in latency-sensitive applications. Instead of optimizing the splatting pipeline itself, we propose \textbf{3DGS3}, a unified post-rendering framework that jointly performs super sampling and frame interpolation through differentiable processing of low-resolution outputs to achieve both high-resolution and high-frame-rate rendering. Our Gradient- -Aware Super Sampling (GASS) module leverages the continuous differentiability of 3DGS to extract image gradients that guide a GRU-based refinement network to enable high-fidelity super sampling. Furthermore, a Lightweight Temporal Frame Interpolation (LTFI) module based on a compact U-Net-like backbone fuses temporal and differentiable spatial cues from consecutive frames to synthesize temporally coherent intermediate frames. Experiments on public datasets demonstrate that 3DGS3 achieves superior rendering efficiency and visual quality when compared with state-of-the-art methods and remains compatible with existing 3DGS acceleration techniques. The code will be publicly released upon acceptance.