Parallel Computing Zernike Momentsvia Combined Algorithms

Abstract

The purpose of this study is to propose two kinds of thread-level parallelization computing Zernike
moments. Our proposed parallel algorithms are based on the combination of the q-recursive method and the
Prata's method with symmetry by a certain dihedral group. The experiments results show that our proposed
method is fast and accurate. The synchronized parallelization is applicable to greater than 250 order Zernike
moments calculation. The reductive parallelization is suitable for computing Zernike moments order between
10 and 250. For computing all Zernike moments up to order 500, it only requires 3.499 sec. on a quad-core
personal computer for a 512 512 test image. By our proposed method, the normalized mean square error with
less than 500 orders Zernike moments is 0.001435 for image reconstruction, whereas the error rate by qrecursion
method is 0.001866.