Resiliency and robustness of alternative shape-based image retrieval techniques

The shape of an object is an important feature for image and multimedia similarity retrieval. However, as a consequence of uncertainty, shape representation techniques may sometimes work well only in certain environments, and their performance may depend crucially on the quality of the technique used to represent the shapes. In this study, we focus on shape-based object retrieval under various uncertainty scenarios and conduct a comparison study on four techniques. We measure the effectiveness of the similarity retrieval of the four different shape representation methods (in terms of recall and precision) under the following situations: (1) in the presence of noise in the database, (2) when the exact corner points are unknown, and (3) factoring in the human perception of similarity. Our results show that the similarity retrieval accuracy of our method [MBC-TPVAS (Minimum Bounding Circle with Touch-Point Vertex-Angle Sequence)] is better than that of the other methods under uncertainty and discrepancies.