Abstract: 3D surface reconstruction are becoming increasing important in geometric modeling and
related applications such as in computer graphics, 3D GIS, reverse engineering. This paper presents an
algorithm that holds the advantages of both region- growth approaches and Delaunay based on
unorganized point cloud. Separation angle is defined and deduced the related conclusion according to the
nature of the surface axis, which is applied to extract triangles from the surface of model and increase the
initial growth area of the region. An approach of adaptive search-ball method is presented to speed up
searching the neighbourhood-triangles and identify the surface boundary. Compared with the traditional
Delaunay-based approach, this algorithm requires only one-pass Delaunay computation and reconstruct
surfaces rapidly without calculation of pole and the vector. Compared with the traditional region growing
method, this algorithm inherits the structural characteristics of the Delaunay triangulation with fewer holes
and accurate the 3D geometry information and topology. Experimental results shows that it is highly 
efficient compared with other existing algorithms and capable of handling surfaces with complex topology,
boundaries, which holds the advantages of both region- growth approaches and Delaunay.

Key words: surface-reconstruction, point-cloud, Delaunay, separation angle, reigon-growing