High Dynamic Range Novel View Synthesis with Single Exposure

TL;DR

This work tackles HDR-NVS when only single-exposure LDR training data are available, addressing motion artifacts and high capture costs of multi-exposure approaches. It introduces Mono-HDR-3D, a meta-algorithm that learns an LDR 3D scene model and lifts it to HDR via camera-imaging aware L2H-CC, while a latent H2L-CC closed loop enables supervision without HDR ground-truth. The framework can be plugged into NeRF or 3DGS representations and is shown to outperform state-of-the-art HDR-NVS methods on synthetic data and maintain strong performance in LDR views, with ablations validating the importance of the camera priors and the closed-loop design. By bridging LDR and HDR spaces through a principled, texture-preserving pipeline, this approach broadens access to HDR scene reconstruction and suggests a practical path toward robust HDR 3D rendering in dynamic or resource-constrained settings.

Abstract

High Dynamic Range Novel View Synthesis (HDR-NVS) aims to establish a 3D scene HDR model from Low Dynamic Range (LDR) imagery. Typically, multiple-exposure LDR images are employed to capture a wider range of brightness levels in a scene, as a single LDR image cannot represent both the brightest and darkest regions simultaneously. While effective, this multiple-exposure HDR-NVS approach has significant limitations, including susceptibility to motion artifacts (e.g., ghosting and blurring), high capture and storage costs. To overcome these challenges, we introduce, for the first time, the single-exposure HDR-NVS problem, where only single exposure LDR images are available during training. We further introduce a novel approach, Mono-HDR-3D, featuring two dedicated modules formulated by the LDR image formation principles, one for converting LDR colors to HDR counterparts, and the other for transforming HDR images to LDR format so that unsupervised learning is enabled in a closed loop. Designed as a meta-algorithm, our approach can be seamlessly integrated with existing NVS models. Extensive experiments show that Mono-HDR-3D significantly outperforms previous methods. Source code will be released.

Figures (8)

• Figure 1: Examples of (a, b) underexposure and (c, d) overexposure. $\Delta t$: Exposure time.
• Figure 2: Overview of Mono-HDR-3D. (a) Given single exposure LDR training images with camera poses, we learn an LDR 3D scene model (e.g., NeRF or 3DGS). (b) Importantly, this LDR model is lifted up to an HDR counterpart via a camera imaging aware LDR-to-HDR Color Converter (L2H-CC). (c) Further, a closed-loop design is formed by converting HDR images back to LDR counterparts with a latent HDR-to-LDR Color Converter (H2L-CC). This enables optimizing the HDR model even with LDR training images, particularly useful in case of no access to HDR training data. During inference, only the HDR or LDR 3D scene model is needed, taking the novel camera view as the input and outputting the corresponding image rendering.
• Figure 3: Structure of our camera imaging aware LDR-to-HDR Color Converter (L2H-CC). $c_i^l$/$c_i^h$: LDR/HDR color; LO: Linear Operation, R: ReLU, SP: Softplus. $\odot$ and $\oplus$: Element-wise multiplication and addition.
• Figure 4: Structure of our camera imaging aware HDR-to-LDR Color Converter (H2L-CC). $\boldsymbol{I}^h$/$\boldsymbol{I}^l$: HDR/LDR image; LO: Linear Operation, R: ReLU, T: Tanh, SM: Sigmoid. $\oplus$: Element-wise addition.
• Figure 5: Comparison of HDR NVS on both (a/b) synthetic and (c) real datasets. $\Delta t$: Exposure time.