NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis
NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis
复制标题
DOI：
10.1145/3503250
复制
发表时间：
2022-01-01
影响因子：
22.7
通讯作者：
Ng, Ren
中科院分区：
文献类型：
--
作者：
Mildenhall, Ben;Srinivasan, Pratul P.;Ng, Ren
我们提出了一种方法，实现国家的最先进的结果，通过优化一个基本的连续体积场景功能，使用稀疏的输入视图的复杂场景的新的意见合成。我们的算法使用完全连接（非卷积）的深度网络表示场景，其输入是单个连续的5D坐标（空间位置（x，y，z）和观察方向（theta，phi）），其输出是该空间位置的体积密度和视图相关的发射辐射。我们合成视图查询5D坐标沿着相机射线，并使用经典的体绘制技术的输出颜色和密度投影到一个图像。因为体绘制是自然可微的，所以优化我们的表示所需的唯一输入是具有已知相机姿势的一组图像。我们描述了如何有效地优化神经辐射场，以渲染具有复杂几何形状和外观的场景的照片级逼真的新视图，并展示了优于神经渲染和视图合成的先前工作的结果。
We present a method that achieves state-of-the-art results for synthesizing novel views of complex scenes by optimizing an underlying continuous volumetric scene function using a sparse set of input views. Our algorithm represents a scene using a fully connected (nonconvolutional) deep network, whose input is a single continuous 5D coordinate (spatial location (x, y, z) and viewing direction (theta, phi)) and whose output is the volume density and view-dependent emitted radiance at that spatial location. We synthesize views by querying 5D coordinates along camera rays and use classic volume rendering techniques to project the output colors and densities into an image. Because volume rendering is naturally differentiable, the only input required to optimize our representation is a set of images with known camera poses. We describe how to effectively optimize neural radiance fields to render photorealistic novel views of scenes with complicated geometry and appearance, and demonstrate results that outperform prior work on neural rendering and view synthesis.