Efficient computation of radial moment functions using symmetrical
property
Pattern Recognition (PR), 39 (11):
2036-2046 (2006)Abstract
The applications of radial moment functions such as orthogonal Zernike
and pseudo-Zernike moments in real-world have been limited by the
computational complexity of their radial polynomials. The common
approaches used in reducing the computational complexity include
the application of recurrence relations between successive radial
polynomials and coefficients. In this paper, a novel approach is
proposed to further reduce the computation complexity of Zernike
and pseudo-Zernike polynomials based on the symmetrical property
of radial polynomials. By using this symmetrical property, the real-valued
radial polynomials computation is reduced to about one-eighth of
the full set polynomials while the computation of the exponential
angle values is reduced by half. This technique can be integrated
with existing fast computation methods to further improve the computation
speed. Besides significant reduction in computation complexity, it
also provides vast reduction in memory storage.
