Organizations: University of Tokyo, Japan. · Tohoku University, Japan. · University of California at Merced and Google DeepMind, US. · University of Tokyo and RIKEN AIP, Japan.
High-quality image acquisition in real-world environments remains challenging due to complex illumination variations and inherent limitations of camera imaging pipelines. These issues are exacerbated in multi-view capture, where differences in lighting, sensor responses, and image signal processor (ISP) configurations introduce photometric and chromatic inconsistencies that violate the assumptions of photometric consistency underlying modern 3D novel view synthesis (NVS) methods, including Neural Radiance Fields (NeRF) and 3D Gaussian Splatting (3DGS), leading to degraded reconstruction and rendering quality. We propose Luminance-GS++, a 3DGS-based framework for robust NVS under diverse illumination conditions. Our method combines a globally view-adaptive lightness adjustment with a local pixel-wise residual refinement for precise color correction. We further design unsupervised objectives that jointly enforce lightness correction and multi-view geometric and photometric consistency. Extensive experiments demonstrate state-of-the-art performance across challenging scenarios, including low-light, overexposure, and complex luminance and chromatic variations. Unlike prior approaches that modify the underlying representation, our method preserves the explicit 3DGS formulation, improving reconstruction fidelity while maintaining real-time rendering efficiency.
Figures & tables
Fig. 1: (a) An overview of Luminance-GS++ , designed to address multi-view images affected by real-world illumination and color degradations via joint global and local adjustments, enabling robust 3DGS reconstruction. (b) Performance comparison between Luminance-GS++ and state-of-the-art methods [ 36 , 28 , 34 , 37 ] under diverse illumination and color conditions.
Fig. 2: Progression from 3DGS [ 6 ] to Luminance-GS [ 36 ] and Luminance-GS++, highlighting joint optimization of input images and 3DGS attributes through global adjustment and local refinement. Multi-view rendering examples illustrate the resulting improvements.
Fig. 3: A low-light “chair” scene from the LOM dataset [ 28 ] . Our global adjustment enhances overall illumination, while our local refinement improves fine details and the quality of pseudo label Cout .
Fig. 4: An overview of the Luminance-GS++ pipeline . Up : Joint optimization of 3D Gaussians with dual color attributes ci and ciout to render input images Cin and pseudo-enhanced outputs Cout . Down : Pseudo-enhancement is achieved through global adjustment (view-adaptive color matrix mapping and curve adjustment; stages (I)–(III)) followed by local refinement using a pixel-wise residual map.
Fig. 5: An illustration of our spatial consistency loss function Lspa , and comparisons with the spatial consistency loss in Aleth-NeRF [ 28 ] .
Fig. 6: (a) Per-view color matrix mapping with learnable transformation Mk . (b) Architecture of the view-adaptive curve generator, which predicts curve bias Lkb from input image Ckin and camera pose. (c) Structure of the view-adaptive parameter generator.
Fig. 7: Effect of varying color temperatures (cool, warm, and mixed) on NVS. We compare the results with and without the color correction loss Lcc and visualize the corresponding color distributions of rendered novel views using scatter plots.
Fig. 8: Chromaticity consistency visualization in CIELAB color space. Pixel chromaticity coordinates ( a∗,b∗ ) from rendered views are plotted for a static region. Luminance-GS++ produces a compact and concentrated distribution, indicating improved cross-view color consistency and stable photometric appearance.
Fig. 9: Efficiency comparison: FPS and peak GPU memory (GB) on the LOM [ 28 ] and Mip-NeRF 360 [ 5 ] datasets. Methods closer to the upper-left corner indicate better efficiency.
Fig. 10: Experimental results on the low-light and overexposure subsets of LOM dataset [ 28 ] . The first column shows the results of the basic 3DGS [ 6 ] , while the second and third columns present the results of various comparison methods [ 28 , 29 , 74 , 20 , 19 , 14 ] . The last two columns show the results of our Luminance-GS [ 36 ] and Luminance-GS++. Some regions are zoomed in to better observe the details.
Fig. 11: Qualitative comparisons of NVS results under lightness, color, and mixed degradation. The last column presents our Luminance-GS++, while the remaining columns correspond to competing methods [ 6 , 34 , 31 , 36 , 37 ] . Zoomed-in regions are provided to highlight details.
Fig. 12: Visualization of the effectiveness of different loss components.
Reconstructing 3D scenes under real-world low-light conditions remains challenging due to severe sensor noise, low signal-to-noise ratios, and degraded photometric consistency, which destabilize geometry estimation and novel view synthesis. Existing approaches often rely on well-lit reference data for reliable Structure-from-Motion (SfM) initialization under degraded inputs or apply per-view enhancement methods that introduce cross-view inconsistencies. To address these limitations, we propose NOVA-GS, a unified noise-aware framework for low-light 3D Gaussian Splatting that subsumes enhancement, denoising, and geometry optimization within a single process. Our method leverages VGGT-based feed-forward estimation to obtain robust camera poses and geometry directly from degraded inputs, eliminating the need for SfM. Building on this initialization, NOVA-GS integrates three coupled components: a structure-aware enhancement module for exposure correction, a self-supervised denoising module with blind-spot masking for pseudo-supervision, and a consistency-driven Gaussian Splatting optimization enforcing cross-view geometric coherence. We further introduce a noise-guided spherical harmonic regularization to suppress view-dependent artifacts in noisy regions. Extensive experiments on diverse real-world low-light datasets demonstrate improved geometric fidelity, color consistency, and robustness without requiring paired supervision or well-lit references. https://shaurya2524.github.io/nova-gs/
Shaurya Pavan A, Vemunuri Divya Madhuri, Yash Pradeep Gawande +1
Novel View Synthesis (NVS) methods, such as 3D Gaussian Splatting (3DGS), rely heavily on the assumption of clean, multi-view consistent, posed input images. Real-world captures can violate this assumption due to screen-space artifacts-static occlusions fixed to the 2D image plane rather than to the 3D world. Common examples include physical sensor defects, environmental obstructions (such as rain or mud on the lens enclosure), capture obstructions (such as a thumb over the camera sensor or a dashboard visible in dashcam footage), and digital overlays (such as watermarks or UI elements). When present, they are erroneously baked into the 3D geometry as "floaters" or near-camera artifacts, degrading the quality of novel-view rendering. In this work, we propose SSA-3DGS, an unsupervised framework that jointly optimizes a 3D scene and a learnable 2D overlay to recover a clean 3D scene and the corrupting artifacts. By exploiting geometric consensus across views, our method effectively disentangles static artifacts from the 3D scene geometry without supervision or manual input. Across diverse synthetic corruptions and a self-captured real-world dataset, SSA-3DGS improves reconstruction fidelity by up to 9 dB PSNR over 3DGS trained on the same corrupted inputs, while faithfully preserving the corrupting artifact.
Kristof Overdulve, Lode Jorissen, Nick Michiels
Digital Future Lab, Flanders Make, Hasselt University
Existing 3DGS methods effectively render high-quality novel views in clear-day scenes. However, they struggle with night scenes, particularly in glow regions, due to the lack of structural features such as textures and edges, which are key cues for splatting-based reconstruction. To address this problem, we leverage a diffusion model and a Vision Foundation Model (VFM) to compensate for missing structural cues. Our method consists of two key novel ideas: semantic feature generation and novel-view semantic learning. First, semantic feature generation produces high-quality semantic features as implicit structural cues for novel views. Specifically, a diffusion model synthesizes novel views with unknown camera poses from training views, while a VFM evaluates their quality. Once high-quality novel views are identified, the VFM extracts robust features to construct the semantic feature bank. Second, novel-view semantic learning enables 3DGS to optimize rendered novel views without requiring ground truth. It achieves this by extracting semantic features from a rendered novel view, searching the feature bank for the most similar features, and minimizing their distance. This process enforces implicit structural constraints, ensuring semantically coherent, artifact-free rendered views. Extensive experiments demonstrate the effectiveness of our GlowGS in generating semantically accurate 3D views, showing significant improvements over existing methods.