Content-based 3D model retrieval using a single depth image from a low-cost 3D camera

Abstract

In this paper, we propose an efficient method for 3D model retrieval using a single depth image. Unlike existing algorithms that use a complete 3D model or a user sketch as input queries, a single depth image is used as an input query, which can be captured easily with an off-the-shelf lowcost 3D camera, such as a Kinect camera. 3D models in the database are represented by multiple depth images acquired from adaptively sampled viewpoints. The proposed algorithm can retrieve relevant 3D models while considering local 3D geometric characteristics using a rotation-invariant feature descriptor. The proposed method consists of three steps: preprocessing, multiple depth image based representation (M-DIBR), and description of 3D models, and similarity measurement and comparison. Experimental results demonstrate that the proposed algorithm is convenient to use and its performance is comparable to recent algorithms in terms of retrieval accuracy and speed.