A new local feature extraction approach for content-based 3D medical model retrieval using shape descriptor

Three-dimensional models are being extensively used in our society. Global shape descriptors are more frequently used in the Content-Based Image Retrieval (CBIR) context due to their robustness and easy implementation, but this kind of descriptor is not adequate for retrieval models with specific characteristics. In this paper a local descriptor is proposed, which analyzes the 3D model shape in different locations of the object in order to increase the retrieval accuracy. Our method is compared with a global descriptor, Distance Histogram, using generic models and specific models which have shape deformations in specific areas. Results show that our method presented higher performance in both contexts.