Content-Based 3D Object Retrieval

Reviewer: Aris Gkoulalas-Divanis

Multimedia repositories have grown rapidly over the past decade. Among the different types of multimedia, the three-dimensional (3D) object is one of the most prevalent data types, having numerous applications. A key issue in this area of research is the retrieval of 3D content, based on user-supplied queries. In order for the database engine to be capable of retrieving accurate content, a measure of similarity must be defined, so that only those objects that are found to match the user query are retrieved from the repository. In this paper, the authors discuss two modern approaches for the measurement of similarity in the retrieval of 3D objects. The first approach, geometric similarity, quantifies the similarity between two 3D objects by measuring the cost of transforming one into the other. The second approach, descriptor-based similarity, first extracts a specific set of numerical features from each 3D object to construct a feature vector representative of the object, and then uses traditional information retrieval to quantify the similarity of the objects, based on the similarity of their feature vectors. The authors present the different steps in the process of extracting a feature vector from a 3D object. Then, they present in detail the two essential properties of a 3D object search engine: efficiency and effectiveness. Finally, they highlight a set of methodologies for the qualitative evaluation of 3D object search engines. Overall, the paper offers a good introduction to the area of content-based 3D object retrieval. It properly motivates the subject area and discusses several open issues and research challenges that have to be more thoroughly addressed in the future. Online Computing Reviews Service