This project is about simplifying a dense, coarse and noisy triangular mesh model while preserving its important features such as places with high curvatures (ridges and ravines). One typical input model is a 3D mesh reconstructed from RGB-D data scanned on a real scene by a consumer depth camera such as Kinect. The output mesh usually contains only 1% or lower number of faces of input mesh. This is done during my internship in Paracosm in Spring 2014.

Refer to the manual file for details about the usage.

An example of clusters in simplification:

Wireframe of final simplified mesh: