The surface density criterion and its application to linear/circular boundary detection and planar/spherical surface approximation

The authors introduce a validity measure called surface density, and develop an unsupervised shell clustering algorithm for the detection of circles in intensity images and for spherical approximation of surfaces in range images. The algorithm can be used when the number of clusters is unknown. Unlike existing algorithms, the new algorithms can handle the degenerate case of lines and planes, and are more robust in the presence of noise