Point-Based Neural Rendering with Per-View Optimization

TL;DR

This paper presents a novel point-based neural rendering approach that improves view synthesis quality and speed by optimizing scene properties with a differentiable pipeline initialized from MVS data.

Contribution

The authors introduce a differentiable point-based rendering pipeline with bi-directional splatting and probabilistic depth testing, enabling better scene optimization and faster rendering.

Findings

Outperforms previous methods in quality and speed across tested scenes.

Effective for multi-view harmonization and stylization.

Allows scene property optimization beyond initial MVS reconstructions.

Abstract

There has recently been great interest in neural rendering methods. Some approaches use 3D geometry reconstructed with Multi-View Stereo (MVS) but cannot recover from the errors of this process, while others directly learn a volumetric neural representation, but suffer from expensive training and inference. We introduce a general approach that is initialized with MVS, but allows further optimization of scene properties in the space of input views, including depth and reprojected features, resulting in improved novel-view synthesis. A key element of our approach is our new differentiable point-based pipeline, based on bi-directional Elliptical Weighted Average splatting, a probabilistic depth test and effective camera selection. We use these elements together in our neural renderer, that outperforms all previous methods both in quality and speed in almost all scenes we tested. Our pipeline can be applied to multi-view harmonization and stylization in addition to novel-view synthesis.