Existing Gaussian splatting methods often fall short in achieving satisfactory novel view synthesis in driving scenes, primarily due to the absence of crafty design and geometric constraints for the involved elements. This paper introduces a novel neural rendering method termed Decoupled Hybrid Gaussian Splatting (DHGS), targeting at promoting the rendering quality of novel view synthesis for static driving scenes. The novelty of this work lies in the decoupled and hybrid pixel-level blender for road and non-road layers, without the conventional unified differentiable rendering logic for the entire scene, while still maintaining consistent and continuous superimposition through the proposed depth-ordered hybrid rendering strategy. Additionally, an implicit road representation comprised of a Signed Distance Field (SDF) is trained to supervise the road surface with subtle geometric attributes. Accompanied by the use of auxiliary transmittance loss and consistency loss, novel images with imperceptible boundary and elevated fidelity are ultimately obtained.
Substantial experiments on the Waymo dataset prove that DHGS outperforms the state-of-the-art methods.
The left part shows the constraints guided by SDF. The top row (right part) displays rendered image and ellipsoids without SDF regularizer, and the bottom row (right part) showcases results obtained with SDF regularizer. It can be observed that the inclusion of SDF regularization leads the road model to render higher-quality images with the help of better road geometry.
The diagram illustrates the proposed depth-ordered hybrid rendering strategy for the environment and road model. Corresponding primitives of each model generate pixels with colors independently through Gaussian splatting. These colors are then composited based on their rendered depths and transmittances, producing the final rendered image.
Comparison of different methods on the Waymo dataset, the left column and right column display the quality of scene reconstruction and novel view synthesis respectively. Our method achieves high-quality reconstruction for both the environment region and the road areas, excelling over other comparative methods in these aspects.
Visual comparisons on the free-view novel view synthesis. The left and right columns respectively exhibit the results of Set3 and Set4 under the free viewpoint setting, where our method significantly outperforms other comparative methods in capturing both road and environment details.
The top row represents the camera layouts after applying various camera pose settings from a bird's eye view, while the bottom row shows the same transformations from a side view.
Visual comparisons on the free-view novel view synthesis in dynamic driving scenes. The 1st, 2nd and 3rd rows respectively exhibit the results of Ground Truth of reconstructing view, Set4 perspective view of Street Gaussian and Set4 perspective view of our method, where our method outperforms Street Gaussian in capturing road details.
The 1st, 2nd and 3rd rows respectively exhibit the results of Ground Truth of reconstructing view, Street Gaussian and our method under free-view novel view synthesis, where our method outperforms Street Gaussian in capturing road details.
