Feature-Preserving Point Cloud Simplification Based on Adaptive Graph Filtering

The simplification of point clouds holds great significance because it decreases computational and storage burdens. Due to the distinct feature of different point cloud data (PCD), most existing resampling methods often need manual adjustments to hyperparameters. To improve the adaptability of simplification, this work proposed a feature-preserving point cloud simplification algorithm based on an adaptive graph filter design. Automating filter coefficient estimation is a significant obstacle, given that different PCDs require adjustments on varying frequencies. Tackling the unreliability of PCD, graph spectral distribution estimation, and adaptive coefficient selection are employed in graph filter design. The proposed method achieves feature-balanced and feature-enhancing PCD simplification based on concise formulations, overcoming the need for manual configuration. Experimental validation on real-world PCDs demonstrates the effectiveness of the proposed method, showcasing competitive performance compared with other established techniques.