New shape descriptor for trademarks retrieval

Summary form only given. This paper introduces a new contour-based shape descriptor for both simple and complex trademark retrieval. The proposed approach not only overcomes the conventional limitation of a contour-based shape descriptor but also outperforms the conventional descriptors. In the paper, first, we use the well-known morphological operation to extract the boundary points from a binary trademark. Then, a sorting labeling method is applied for computing the similarity. Experimental results show that the proposed two methods are robust and reduce the complexity of extracting features significantly.