Distance Difference for Fast Shape Recognition to Minimum Enclosing Polygons

This paper presents a fast shape recognition method based on the Distance Difference between internal and Euclidean distances within a Minimum Enclosing Polygon(DD-MEP). The core idea is to first compute the minimum enclosing polygon of a shape, and then extract a saliency-constrained distance difference feature by comparing the distance from the shape's centroid to contour points (internal distance) and to the polygon boundary (Euclidean distance). By extracting features only from salient regions inside the polygon, the method improves both robustness and computational efficiency. A histogram-based representation is used to describe the distribution of distance differences across the shape contour. Experiments conducted on MPEG-7 Part $B$ and Swedish Leaf datasets demonstrate that the proposed method outperforms several classical shape descriptors in terms of recognition accuracy and speed.