This paper describes an automatic 3D surface scanning technique using a 3D scanner. It allows the acquisition of a complete surface model of a 3D artifact without any manual registration and human interference. A two-pass approach is applied using a rotary table. In the first pass, a sequence of 2D images of the artifact are collected using a small rotation step. An image similarity measure is then taken to compare adjacent images to assess the differences between consecutive images to establish the optimal scanning angles. In the second pass, 3D scans are taken using the scanning angles derived in the first pass. An Iterative Closest Point (ICP) algorithm is employed to provide precise data alignment of neighboring scans. A new space encoding algorithm is developed to filter the over sampling areas for smoother polygonization. This technique is being implemented in a digital library project for the reconstructions of 3D computer models for artifacts from a collection in several Museums.