Hierarchical frequency method compresses large point clouds efficiently
Hierarchical Frequency-Domain Compression of Implicit Geometric Representations for Large-Scale Point Clouds
Computer Vision and Pattern Recognition
Summary
Large 3D point cloud models can be too big to handle easily on computers. This paper introduces a way to store these models more compactly by breaking down the shapes into simpler parts based on frequencies, like sound waves. The method keeps basic shapes clearly and adds details carefully so the model looks right after being compressed. The researchers showed their method works well on huge 3D scans with millions of points, shrinking the size more than older methods without losing important detail.
What this means in practice
- •For 3d graphics engineers: Store and transmit large 3D point cloud scans with less memory while preserving visual detail using frequency-based hierarchical compression.
- •For geospatial data teams: Manage and compress complex large-scale geographic 3D point clouds efficiently for mapping and analysis tasks.
Authors
Manlin Yao, Jiabin Liu, Guan Wang, Haixu Liu, Hui Li
Abstract
Large-scale point cloud representations of complex geome tries incur prohibitive computational and memory costs, necessitating compressed implicit representations. To ad dress this, we propose a unified framework comprising im plicit geometric field representation, hierarchical frequency domain compression, and conditional high-frequency predic tion. Specifically, an unordered point cloud is mapped to an implicit field defined within its physical bounding box. A smooth Fourier pyramid is then constructed, where com pact low-frequency components capture the global geometry. Inter-scale high-frequency residuals are encoded to preserve the spatial information required for reconstructing fine geo metric details. To restore the high-frequency information lost during compression, we develop a hierarchical 3D neural net work. The reconstructed implicit field is converted back into a point cloud through isosurface extraction. Experiments on a complex-boundary point cloud with more than eight mil lion points demonstrate that the proposed method achieves a higher compression ratio than existing point cloud compres sion methods while maintaining comparable reconstruction quality.