View-dependent triangle mesh simplification using GPU-accelerated vertex removal

Abstract

We present an approach to view-dependent triangle mesh simplification based on vertex removal, which focuses on allowing the execution of a large number of operations in parallel. The individual vertex removal operations are designed to be applied without any need for communication or synchronisation between operations, thus allowing an efficient implementation on modern GPUs to reduce the computation time for the coarse mesh. Since we cannot compute the entire simplification in a single step and have to perform several iterations of parallel vertex removal, we aim to maximize the number of vertices removed from the mesh in each iteration to efficiently use the available hardware and reduce the number of necessary iterations. The removal operation is based on the half edge collapse and avoids mesh foldovers and topological inconsistencies at each step.