A Novel Shape signature of multiply-connected domain
Project Description:
The study of 2D shapes is a central problem in many different research
areas, such as computer vision and medical imaging. In 2D shape analysis,
classification and recognition of objects from their observed sihouette
are extremely crucial and yet difficult. It usually involves a defintion
of a metric on the 2D shape space, so that its mathematical structure can
be used for further analysis. Although significant progress can be found
for the study of 2D simply-connected shapes, none or very little works have
been done on the study of 2D multiply-connected domains. In this work, we
proposed a representation of 2D multiply-connected domains using conformal
geometry. A natural metric can be defined on the proposed representation
space. Hence, a shape signature can be defined to measure the similarities
between objects. This is done by mapping the exterior and interior of the
object conformally to unit disks and punctual disks. A set of diffeomorphisms
from the unit disks to itself can be obtained, which are used to define
shapes. We prove mathematically that our proposed shape signature represents
shape up to scaling and translation. Experimental results shows the effectiveness
of our propsed method for 2D shape analysis.
Publication:
Result: