Point Cloud Rendering via Multi-plane NeRF

We propose a new neural point cloud rendering method by combining point cloud multi-plane projection and NeRF [12]. Existing point-based rendering methods often rely on the high-quality geometry of point clouds. Meanwhile, NeRF and its extensions usually query the RGB and volume density of each point through neural networks, thus leading to a low inference efficiency. In this paper, we project point features to multiple random depth planes and feed them into a 3D convolutional neural network to predict the RGB and volume density maps. Then we synthesize a novel view through volume rendering. Projecting point features to multiple planes reduces the impact of geometry error and improves the rendering efficiency. Experimental results on the DTU and ScanNet dataset show that our approach achieves state-of-the-art results. Our source code is available at https://github.com/Mayxmu/PCMP-NeRF .