3D Models Retrieval based on Shape-Similarity

Abstract

Richer types of multimedia such as audio, video and 3D objects are becoming more and more common place. However, current retrieval techniques in these areas are not as sophisticated as textual and 2D image techniques and in many cases rely upon textual searching through associated keywords. The work presented in this paper aims to retrieve similar 3D shapes from the large number of 3D object datasets. Firstly three virtual views, top view, side view and front view, of an object are extracted from 3D objects to perform dimension reduction of 3D model into 2D model. Secondly, we extract the features of the shape by calculating the ratios of virtual views of an object using Euclidean distance. For searching shape similarity, k-nearest neighbor algorithm is applied. To demonstrate the efficiency of proposed system regarding the computational aspects and retrieval performance, we tested the proposed system on NTU 3D model database.