The introduction of the Princeton Shape Benchmark database (PSB) [Shilane et al. 2004] has given rise to extensive research into ways to accurately perform 3D matching. Among the extant methods proposed, the LightField Descriptor (LFD) [Chen et al. 2003] based method currently is the most accurate, but it suffers from a deficiency in on-line matching speed. The contribution of this research is twofold. First, fewer characteristic views are used instead of huge images to boost the on-line retrieval time. Second, depth information is employed to keep the retrieval accuracy.