Radiance field methods represent the state of the art in reconstructing complex scenes from multi-view photos. However, these reconstructions often suffer from one or both of the following limitations: First, they typically represent scenes in low dynamic range (LDR), which restricts their use to evenly lit environments and hinders immersive viewing experiences. Secondly, their reliance on a pinhole camera model, assuming all scene elements are in focus in the input images, presents practical challenges and complicates refocusing during novel-view synthesis. Addressing these limitations, we present a lightweight method based on 3D Gaussian Splatting that utilizes multi-view LDR images of a scene with varying exposure times, apertures, and focus distances as input to reconstruct a high-dynamic-range (HDR) radiance field. By incorporating analytical convolutions of Gaussians based on a thin-lens camera model as well as a tonemapping module, our reconstructions enable the rendering of HDR content with flexible refocusing capabilities. We demonstrate that our combined treatment of HDR and depth of field facilitates real-time cinematic rendering, outperforming the state of the art.

## Related Work

### Cinematic Gaussians: Real-Time HDR Radiance Fields with Depth of Field

This paper builds on previous research in several areas related to rendering high dynamic range (HDR) scenes with realistic depth of field effects. Some key related works include:

- [HDR Gaussians: Efficient High Dynamic Range Radiance Fields](https://aimodels.fyi/papers/arxiv/hdr-gs-efficient-high-dynamic-range-novel): This paper introduced the use of Gaussian primitives for efficient HDR radiance field rendering.
- [Lighting Every Darkness: 3D Gaussian Splatting for Fast Real-Time Global Illumination](https://aimodels.fyi/papers/arxiv/lighting-every-darkness-3dgs-fast-training-real): This work demonstrated how 3D Gaussian splatting can enable fast global illumination calculations.
- [2D Gaussian Splatting: Geometrically Accurate Radiance Fields from Sparse Samples](https://aimodels.fyi/papers/arxiv/2d-gaussian-splatting-geometrically-accurate-radiance-fields): This paper explored using 2D Gaussian splatting to reconstruct accurate radiance fields from sparse samples.
- [Deblurring 3D Gaussian Splatting](https://aimodels.fyi/papers/arxiv/deblurring-3d-gaussian-splatting): This work addressed the issue of blur artifacts in 3D Gaussian splatting by introducing a deblurring technique.
- [3D Gaussian Splatting: Deferred Reflection](https://aimodels.fyi/papers/arxiv/3d-gaussian-splatting-deferred-reflection): This paper demonstrated how 3D Gaussian splatting can be used for efficient deferred reflection calculations.

The current paper, "Cinematic Gaussians," aims to build on these previous advancements by developing a real-time HDR radiance field rendering system that also supports realistic depth of field effects.