Efficient and prioritized point subsampling for CSRBF compression

Abstract

We present a novel cost function to prioritize points and subsample a point set based on the dominant geometric features and local sampling density of the model. This cost function is easy to compute and at the same time provides rich feedback in the form of redundancy and non-uniformity in the sampling. We use this cost function to simplify the given point set and thus reduce the CSRBF (Compactly Supported Radial Basis Function) coefficients of the surface fit over this point set. Further compression of CSRBF data set is effected by employing lossy encoding techniques on the geometry of the simplified model, namely the positions and normal vectors, and lossless encoding on the CSRBF coefficients. Results on the quality of subsampling and our compression algorithms are provided. The major advantages of our method include highly efficient subsampling using carefully designed, effective, and easy compute cost function, in addition to a very high PSNR (Peak Signal to Noise Ratio) of our compression technique relative to other known point set subsampling techniques.