Distributed Reconstruction of Implicit Surfaces

This paper presents an approach to use the partition of unity technique associated with the radial basis technique to provide implicit surface reconstruction over a distributed environment. The proposed algorithm acquires all the advantages of the partition of unity with better performance and capability of reconstructing surfaces from larger data sets. The global large implicit surface reconstruction problem is partitioned into smaller independent radial basis functions fitting problems. These local problems are distributed over the processing nodes solved independently then the recombined together over the master processing node using the partition of unity. The experimental results for the benchmark models demonstrated that the proposed algorithm can reduce the total radial basis functions fitting time with about 66% using 10 processing nodes. Also, the large models with size of O(106) points can be reconstructed with acceptable reconstruction time.