New approach for efficiently retrieving similar 3D models based on reducing the research space

The need of efficient methods for 3D shape retrieval that satisfies simultaneously the computational efficiency (Fastness) and the quality of the retrieval results (Relevance), is an active topic in various research communities. Designing such methods is a great challenge. In order to satisfy the computational efficiency without affecting the performance, we propose in this paper a novel approach that adjusts the compromise ”Fastness/Relevance”. Our key idea is to reduce the research space during the retrieval process, by ignoring 3D-objects that are not similar to the query. To do this, we propose to cooperate two existing methods. The first one satisfies the condition of the fastness while the second one satisfies the condition of the relevance. First, we use the fastness method to reduce the research space. Then, the second method is used to perform the retrieval in the reduced research space. For experimental tests, we have applied our approach to BF-SIFT and CM-BOF methods. The obtained experimental results, on the PSB database, show that our approach efficiently adjusts the compromise ”Fastness/Relevance”.