3D Blur Kernel on Gaussian Splatting


Yongchao Lin (Inner Mongolia University), Xiangdong Su (Inner Mongolia University), Yuhan Yang (Inner Mongolia University )
The 35th British Machine Vision Conference
PDFPosterVideo (Right click to download)

Abstract

3D Gaussian Splatting, as a distinct rendering pipeline from Neural Radiance Field (NeRF), has garnered considerable attention in 3D scene reconstruction tasks due to its remarkable performance and efficient training and rendering capabilities. However, it requires high-quality input scene images. Image blurring resulting from defocus or camera motion can significantly degrade the reconstruction quality, leading to artifacts when reconstructing images with new viewpoints. To solve this problem, we propose 3D Blur Kernel on Gaussian Splatting, a method to characterize a 3D blur kernel using spherical harmonic functions. Specifically, we optimize a spherical harmonic function to characterize the 3D blur kernel; the blurring process is simulated by convolving the viewpoint-specific blur kernel, formed through projection, with the covariance matrix of 3D Gaussians and by applying transformations and rotations to the 3D Gaussians. Additionally, we introduce a density regularization term to enhance the reconstruction of texture details. We conduct experiments on both real and synthetic data, and the experimental and qualitative results demonstrate the effectiveness of our method.

Citation

@inproceedings{Lin_2024_BMVC,
author    = {Yongchao Lin and Xiangdong Su and Yuhan Yang},
title     = {3D Blur Kernel on Gaussian Splatting},
booktitle = {35th British Machine Vision Conference 2024, {BMVC} 2024, Glasgow, UK, November 25-28, 2024},
publisher = {BMVA},
year      = {2024},
url       = {https://papers.bmvc2024.org/0648.pdf}
}


Copyright © 2024 The British Machine Vision Association and Society for Pattern Recognition
The British Machine Vision Conference is organised by The British Machine Vision Association and Society for Pattern Recognition. The Association is a Company limited by guarantee, No.2543446, and a non-profit-making body, registered in England and Wales as Charity No.1002307 (Registered Office: Dept. of Computer Science, Durham University, South Road, Durham, DH1 3LE, UK).

Imprint | Data Protection