QuantumGS: Quantum Encoding Framework for Gaussian Splatting

TL;DR

QuantumGS presents a quantum–classical hybrid for neural rendering by embedding Variational Quantum Circuits into Gaussian Splatting. It maps viewing directions onto the Bloch sphere using a $3$-qubit encoding with rotations $R_y$ and $R_z$, and uses hypernetworks or hash-conditioned conditioning to generate VQC parameters, enabling per-Gaussian or global residuals to capture high-frequency view-dependent radiance. The dual pipelines—per-Gaussian hyper-quantum modeling and joint-hash global modeling—achieve state-of-the-art fidelity on synthetic data and robust generalization on real-world scenes, while maintaining interactive frame rates. This work demonstrates the viability of quantum-geometric encoding for rendering and points toward hardware quantum acceleration as a pathway to further improvements in photorealistic, view-dependent appearance modeling.

Abstract

Recent advances in neural rendering, particularly 3D Gaussian Splatting (3DGS), have enabled real-time rendering of complex scenes. However, standard 3DGS relies on spherical harmonics, which often struggle to accurately capture high-frequency view-dependent effects such as sharp reflections and transparency. While hybrid approaches like Viewing Direction Gaussian Splatting (VDGS) mitigate this limitation using classical Multi-Layer Perceptrons (MLPs), they remain limited by the expressivity of classical networks in low-parameter regimes. In this paper, we introduce QuantumGS, a novel hybrid framework that integrates Variational Quantum Circuits (VQC) into the Gaussian Splatting pipeline. We propose a unique encoding strategy that maps the viewing direction directly onto the Bloch sphere, leveraging the natural geometry of qubits to represent 3D directional data. By replacing classical color-modulating networks with quantum circuits generated via a hypernetwork or conditioning mechanism, we achieve higher expressivity and better generalization. Source code is available in the supplementary material. Code is available at https://github.com/gwilczynski95/QuantumGS

Figures (7)

• Figure 1: Top: Truck scene from Tanks and Temples knapitsch2017tanks demonstrates complex transparency. Standard 3DGS blurs the poster behind the windshield due to low-frequency spherical harmonics. QuantumGS preserves high-frequency view-dependence, recovering background visibility. Bottom: Directional color response of a single Gaussian. Unlike smooth SH patterns (middle), Bloch-sphere encoding (right) learns complex, irregular responses (e.g., central dark lobe), enabling precise light transmission modeling.
• Figure 2: Our framework integrates quantum processing into 3D Gaussian Splatting for view-dependent color and opacity residuals. Top: Two interchangeable pipelines. Pipeline I (Hyper-Quantum) generates per-Gaussian VQC parameters via hypernetwork from spatial hash encoding. Pipeline II (Joint-Hash Global) feeds spatial and directional hash features to a shared quantum network. Bottom: Hybrid QMLP maps viewing directions to Bloch sphere via rotation gates ($R_y$, $R_z$), processes through VQC with circular entanglement, and decodes measurements via classical MLP to guide rendering.
• Figure 3: Top: The Ship scene (NeRF Synthetic) features thin rigging and liquid transparency. Standard 3DGS produces distracting "floater" artifacts beneath the model. QuantumGS generates a clean background comparable to ground truth. Bottom: In the Room scene (Mip-NeRF 360), standard 3DGS struggles with geometric consistency at the bookshelf base, creating jagged artifacts. QuantumGS eliminates these errors, preserving straight lines and structural coherence.
• Figure 4: Object-centric scenes. In the Drums scene, VDGS blurs the reflection on the drum surface, losing the geometric definition of the reflected drum. QuantumGS preserves the distinct shape of the reflection. In the LEGO scene, VDGS exhibits floater artifacts near the roof. Additionally, QuantumGS recovers occlusion shadows on the chassis, unlike standard 3DGS.
• Figure 5: Comparisons on real-world datasets. In the Truck scene (Tanks and Temples), standard 3DGS fails to capture high-frequency reflections on the windshield, resulting in a blurred appearance, whereas QuantumGS recovers sharp specular details. In the Kitchen scene (Mip-NeRF 360), standard 3DGS renders the LEGO truck with unnatural "foggy" or hazy appearance due to lighting ambiguity. QuantumGS resolves this issue, producing clear object boundaries. In the Dr. Johnson scene (Deep Blending), QuantumGS correctly handles high-dynamic-range light entering through the window, while standard 3DGS produces unnatural overexposure and artifacts on surrounding walls.