Shape Retrieval Algorithm Based on Distance

Abstract—we introduce a new feature vector for shape-based image retrieval. This feature depends on two factors: the centroidal distances of shape and correlation between neighboring edges, hence this scheme is not only effective and robust to translation, scaling, and rotation, but also has the qualities of expressing the spatial information of edges. Experimental results show the proposed method has clearer superiority and higher average precision and recall rates than several other methods. Index Terms—image retrieval; shape representation; distance autocorrelogram; distance histograms I