A global clustering approach to point cloud simplification with a specified data reduction ratio

This paper studies the problem of point cloud simplification by searching for a subset of the original input data set according to a specified data reduction ratio (desired number of points). The unique feature of the proposed approach is that it aims at minimizing the geometric deviation between the input and simplified data sets. The underlying simplification principle is based on clustering of the input data set. The cluster representation essentially partitions the input data set into a fixed number of point clusters and each cluster is represented by a single representative point. The set of the representatives is then considered as the simplified data set and the resulting geometric deviation is evaluated against the input data set on a cluster-by-cluster basis. Due to the fact that the change to a representative selection only affects the configuration of a few neighboring clusters, an efficient scheme is employed to update the overall geometric deviation during the search process. The search involves two interrelated steps. It first focuses on a good layout of the clusters and then on fine tuning the local composition of each cluster. The effectiveness and performance of the proposed approach are validated and illustrated through case studies using synthetic as well as practical data sets.