Quadric surface fitting for sparse range data

The authors present a systematic comparison of three commonly used least-squares based methods that describes the relationship between noise levels, patch sizes and reliability of surface classification in computer vision. The different methods were tested on several sets of synthetic and real data. Complete sets of quadric surfaces were tested. In each case the standard deviation of the Gaussian noise ranged from 0 to 0.05. Four different patch sizes were tested in each case representing data from the entire surface, half surface, quarter surface and from a small patch of the surface. Similar tests were made for two different types of real range data, a sphere and a cylinder. Both synthetic and real data showed improvement in the results obtained by using the M-estimate method in the case of gross outlying points.<
>