DARB-Splatting: Generalizing Splatting with Decaying Anisotropic Radial Basis Functions
This work addresses a computational bottleneck in 3D reconstruction for graphics and vision applications, offering significant efficiency gains, though it is incremental as it builds on existing splatting frameworks.
The paper tackled the limitation of splatting-based 3D reconstruction methods to exponential family kernels by introducing decaying anisotropic radial basis functions (DARBFs), achieving up to 34% faster training convergence and 45% reduced memory consumption while maintaining similar quality metrics like PSNR.
Splatting-based 3D reconstruction methods have gained popularity with the advent of 3D Gaussian Splatting, efficiently synthesizing high-quality novel views. These methods commonly resort to using exponential family functions, such as the Gaussian function, as reconstruction kernels due to their anisotropic nature, ease of projection, and differentiability in rasterization. However, the field remains restricted to variations within the exponential family, leaving generalized reconstruction kernels largely underexplored, partly due to the lack of easy integrability in 3D to 2D projections. In this light, we show that a class of decaying anisotropic radial basis functions (DARBFs), which are non-negative functions of the Mahalanobis distance, supports splatting by approximating the Gaussian function's closed-form integration advantage. With this fresh perspective, we demonstrate up to 34% faster convergence during training and a 45% reduction in memory consumption across various DARB reconstruction kernels, while maintaining comparable PSNR, SSIM, and LPIPS results. We will make the code available.