3D object retrieval and pose estimation for a single-view query image in a mobile environment

3D object retrieval and its pose estimation for a single view query image are essential operations in many specialized applications. With the recent deployment of various mobile devices, such operations require near real-time performance. However, most of the existing methods are not appropriate for mobile devices, due to their massive resource requirements. In this paper, we propose new 3D object retrieval and pose estimation schemes that can be used on a client-server platform. In order to accomplish this, we first construct both a sparse and a full index on the shape feature of the objects for the client and the server, respectively. Then, the client (the mobile device) retrieves the candidate camera view images that are similar to the query image by using the sparse index. The server refines the results by using the full index and then computes the exact pose by using the SIFT (Scale Invariant Feature Transform) features. In the experiment, we show that our prototype system based on the proposed scheme can achieve an excellent performance.