Multi-Frequency-Aware Patch Adversarial Learning for Neural Point Cloud Rendering

TL;DR

This paper introduces a multi-frequency-aware patch adversarial learning framework for neural point cloud rendering, significantly improving realism by aligning spectral and spatial features, and enhancing training stability.

Contribution

The paper proposes a novel multi-discriminator scheme combining spectral and spatial domain discriminators, and a noise-resistant voxelisation method for improved neural point cloud rendering.

Findings

Achieves state-of-the-art rendering quality

Improves convergence speed and stability

Effectively captures spectral distributions of real images

Abstract

We present a neural point cloud rendering pipeline through a novel multi-frequency-aware patch adversarial learning framework. The proposed approach aims to improve the rendering realness by minimizing the spectrum discrepancy between real and synthesized images, especially on the high-frequency localized sharpness information which causes image blur visually. Specifically, a patch multi-discriminator scheme is proposed for the adversarial learning, which combines both spectral domain (Fourier Transform and Discrete Wavelet Transform) discriminators as well as the spatial (RGB) domain discriminator to force the generator to capture global and local spectral distributions of the real images. The proposed multi-discriminator scheme not only helps to improve rendering realness, but also enhance the convergence speed and stability of adversarial learning. Moreover, we introduce a noise-resistant voxelisation approach by utilizing both the appearance distance and spatial distance to exclude the spatial outlier points caused by depth noise. Our entire architecture is fully differentiable and can be learned in an end-to-end fashion. Extensive experiments show that our method produces state-of-the-art results for neural point cloud rendering by a significant margin. Our source code will be made public at a later date.