Shape signature based on Homotopic deformation

摘要

This paper proposes a new method for shape signature which is based on Homotopic deformation(HDBS). HDBS use the minimum circumcircle which encircle the shape as the deform template, and use the path length derived from deformation as the sampling data. This method ensures the disjoint among the paths of deformation, avoids crossing by paths and contours except at origin, every point in contour can be sampled, as a result, HDBS can accurately derive shape signature for the complex concave shapes. Moreover, our approach captures global features of shapes and is less sensitive to uneven noise. Through theoretical prove and experimental test, the signature data is not sensitive to position move, rotation and scale, and can be used in shape match directly.