Abstract—In this paper we focused on designing a novel
hybrid regional descriptor for shape classification, which
includes three novel descriptors regional area descriptor (RAD),
regional skeleton descriptor (RSD) and tangent function(TF).
The RAD and RSD are based on the primitive skeleton of shape
while the tangent function is developed from the contour by
finding out the important landmark points and collecting
regional information around them. In the matching stage, a
customized Optimal Path Searching algorithm is integrated into
our setting as distance measure function, which is an extension
of efficient dynamic programming algorithm. The proposed
shape descriptors are tested on commonly used datasets and the
results are analyzed and compared to state-of-the-art methods.
The experimental results show that, compared to others, our
results are satisfactory.
Index Terms—Shape classification, skeleton, contour, rad,
rsd, tangent function, optimal path searching.
The authors are with Department of Computer and Information Science,
University of Macau, China (e-mail: cmpun@umac.mo).
Cite:Cong Lin and Chi-Man Pun, "Shape Classification Using Hybrid Regional and Global Descriptor," International Journal of Machine Learning and Computing vol.4, no. 1, pp. 68-72, 2014.