|M. G. Strintzis|
|ELSEVIER, Pattern Recognition, Volume 42 , Issue 11, pp. 2447-2459, Nov 2009|
In this paper, the novel 3D Shape Impact Descriptor is introduced, which is based on the resulting gravitational phenomena in the surrounding area of every 3D object. The 3D object is considered as a distributed 3D mass and the descriptor of the 3D object is indirectly computed from the resulting fields. The field is described using both Newton's and General Relativity's laws. In the Newtonian approach, histograms of the field values in the surrounding area of the 3D object are computed, while in the Relativistic approach the descriptors are histograms of the time-space curvature in the surrounding area of the 3D object. The basic motivation behind the proposed approach is the robustness with respect to object's degeneracies and the native invariance of the resulting descriptors under rotation and translation. Experiments which were performed in various 3D object databases proved that the proposed method can be efficiently used for 3D object retrieval applications.