IgnisGrowth: A Novel Combustion-Growth Inspired Method for Point Cloud Completion

Point cloud completion is a key task in 3D point cloud processing, addressing the challenge of reconstructing the full shape of a 3D object from incomplete data. We propose a novel method, IgnisGrowth, which simulates a gradual "growth" process of the point cloud. Inspired by natural growth mechanisms, our method starts from local regions and expands to the global structure, recovering local details and ensuring global consistency. Extensive experiments show that IgnisGrowth outperforms existing methods on widely used benchmarks. The code is available at https://github.com/yizilaohan/IgnisGrowth.