A memory effective two-phase approach for large scanned surface mesh simplification

We present a novel two-phase multi-attribute algorithm suitable for large surface mesh simplification. By employing a linear combination of error metrics to control the process, the proposed algorithm incorporates geometric error control and preserves other attributes of the original model such as the texture (vertex color) and surface normal. In the first phase, we utilize the volume- surface tree [1] (VS-Tree) for vertex clustering to achieve memory effectiveness and computational efficiency. In the second phase, an iterative edge contraction process is applied to obtain the final simplified model. We experiment the proposed algorithm to large mesh models and the results are compared with those from other state of art algorithms such as the octree clustering simplification[5] (OCS) and the original quadric error metric (QEM) based simplification (Q-Slim) ([2][3]).