A fast, memory efficient, linear time algorithm that generates a manifold triangular mesh has been developed [13]. It has been used on standard point cloud data sets as well as the data sets acquired by our prototype range system. The speed of the algorithm is derived from a projection-based approach used to determine the incident faces on a point.

An example of the output from this algorithm is shown in figure 5.
The original input has 6.5 million samples, while the result has only
`~`

140k points. The processing time taken to build the mesh was 88
seconds on an Oynx2.
The rendering of the reconstructed geometric model is made by applying
the intensity image as a texture map onto the created triangle mesh.
Despite the drastic reduction in complexity, the model retains the
important features of the collected data set.