HS-Surf: A Novel High-Frequency Surface Shell Radiance Field to Improve Large-Scale Scene Rendering

Jiongming Qin, Fei LUO, Tuo Cao, Wenju Xu, Chunxia Xiao

Abstract: Prior neural radiance fields often struggle to preserve high-frequency textures in urban and aerial large-scale scenes due to insufficient model capacity on the scene surface. This is attributed to their sampling locations or grid vertices falling in empty areas. Additionally, most models do not consider the drastic changes in distances. To address these issues, we propose a novel high-frequency surface shell radiance field, which uses depth-guided information to create a shell enveloping the scene surface under the current view, and then samples conic frustums on this shell to render high-frequency textures. Specifically, our method comprises three parts. Initially, we propose a strategy to fuse voxel grids and information of distance scales to generate a coarse scene at different distance scales. Subsequently, we construct a shell based on the depth information to carry out compensation to incorporate texture details not captured by voxels. Finally, the smooth and denoise post-processing further improves the rendering quality. Substantial scene experiments and ablation experiments demonstrate that our method achieves the obvious improvement of high-frequency textures at different distance scales and outperforms the state-of-the-art methods.

Primary Subject Area: [Experience] Multimedia Applications

Secondary Subject Area: [Experience] Art and Culture

Relevance To Conference: Rendering novel views on urban or aerial large-scale scenes is a long-standing challenge in computer vision. Neural radiation field (NeRF) provides important technical support for related applications, such as smart cities, virtual reality, and autonomous driving. However, previous NeRFs having a low efficiency in increasing the model capacity on scene surfaces, resulting in insufficient capacity to render high-frequency textures. This defect makes it easy for NeRFs to generate blurry renders in large-scale scenes, because of more empty areas and complex appearances. To this end, the proposed HS-Surf constructs a shell enveloping the scene surface under the current view, and samples conic frustums on this shell to render high-frequency textures. This novel approach confines the rendering to the scene surface to maximize capacity utilization, overcoming the inefficiency in previous methods. The capacity on the surface is efficiently increased, greatly enhancing the rendering quality of high-frequency textures.

Submission Number: 3988