FFTs for the 2-Sphere-Improvements and Variations

作者: D.M. Healy , D.N. Rockmore , P.J. Kostelec , S. Moore

DOI: 10.1007/S00041-003-0018-9

关键词: Fourier analysisFourier transformConvolutionPentiumMathematicsAlgorithmHeuristicsInverseFloating pointPartial differential equation

摘要: Earlier work by Driscoll and Healy has produced an efficient algorithm for computing the Fourier transform of band-limited functions on 2-sphere. In this article we present a reformulation variation original which results in a greatly improved inverse transform, consequent convolution for such functions. All require at most O(N log2 N) operations where N is number sample points. We also address implementation considerations give heuristics allowing reliable computationally floating point implementations slightly modified algorithms. These claims are supported extensive numerical experiments from our implementation C DEC, HP, SGI Linux Pentium platforms. These results indicate that variations both large range of useful problem sizes. Performance appears to be architecture-dependent. The concludes with brief discussion few potential applications.

参考文章(39)
David Keith Maslen, Fast transforms and sampling for compact groups Harvard University. ,(1993)
P. P. Vaidyanathan, Multirate Systems and Filter Banks ,(1992)
Kenichi Kanatani, Group-Theoretical Methods in Image Understanding Springer-Verlag New York, Inc.. ,(1990) , 10.1007/978-3-642-61275-6
Aho AV, JE Hopcroft, JD Ullman, The Design and Analysis of Computer Algorithms ,(1974)
Alan V. Oppenheim, Ronald W. Schafer, Discrete-Time Signal Processing ,(1989)