Research on Shape Representation Based on Statistical Features of Centroid-contour Distance

This paper proposes a novel shape representation method based on statistical features. According to the joint analysis on Centroid-Contour Distance (CCD) and chaincode, the silhouette is decomposed into several levels based on CCD. And then, the chaincode describing laying in each level is analyzed to extract the Joint Statistical of Centroid-Contour Distance and Chaincode (JSCCDC) descriptor for the silhouette. The similarity between different shapes can be measured by the city-block distance. Experiment results show that the proposed method describes both global and local features. Compared with traditional feature weighting method, JSCCDC is more accurate and reliable for shape matching and retrieval.