Based on the metaball models

a practical method is presented for automatically reconstructing an implicit surface from given sample points on the surface of an object in R3. Firstly

a spherical representation of the model from the samples can be obtained. Secondly

the spheres are utilized as initial metaballs—this greatly improves the robustness and efficiency of the metaball-based approximation

and the initial implicit surface is constructed. Lastly

the resulting implicit surface is created by adjusting every meatball parameter in an optimizing process. The optimal spheres and local supported field functions accelerate the surface optimization strongly. Experiments demonstrate the method’s efficiency and robustness.