@article{oai:chukyo-u.repo.nii.ac.jp:00012560,
 author = {輿水, 大和 and KOSHIMIZU, Hiroyasu},
 issue = {1},
 journal = {中京大学教養論叢, Chukyo University bulletin of the Faculty of Liberal Arts},
 month = {Jun},
 note = {Generally, all sepectrum components are simultaneously calculated in two-dimensional fast Fourier transform (2-D FFT) even when only a previously specified subset of them is required. Five typical methods for calculating such a subset of components are introduced and the properties of them and the relations among them are discussed from a viewpoint of the computation costs. One of those methods is effective in comparison with 2-D FFT method when s×t 〓 N. (log_2 (M×N) ? log_2N). (M, N : the size of image data, s×t : the number of specified components, 1〓s〓M, 1〓t〓N) Another method is always effective when s×t varies from 1×1 to M×N. The latter is less effective than the former when s×t〓(log_2 (M+N))/(M×N). The priority measure for the selection of the best method w. r. t. computation time is exactly made clear for 0〓(s×t)/(M×N)〓1., 5, KJ00004205414, 論文, Article},
 title = {成分選択的2次元フーリエ変換法の選択的利用法},
 volume = {28},
 year = {1987},
 yomi = {コシミズ, ヒロヤス}
}