Many image processing applications require an extensive usage of FFT2 routine (or, in the most general case, a N-dimensional FFT) of matrices having the same dimensions. In these cases MATLAB FFT2 can result extremely inefficient. In general the execution time can be significantly reduced by splitting the N-dimensional FFT into several unidimensional FFT. A good trick is to apply the fft operator varying the minimum number of times the length of the unidimensional vectors which have to be FFT-transformed. You might be able to increase the speed of fft using the utility function fftw, which controls how MATLAB optimizes the algorithm used to compute an FFT of a particular size and dimension. In the following examples the planner is always 'hybrid'. The best vectorization strongly depends on the dimensions of input matrices. Chose the optimal solution comparing the execution times of the methods proposed.
Via Centrale 35
67042 Civita Di Bagno
L'Aquila --- ITALY
mobile +39 3207214179
Luigi Rosa (2020). FFT2 optimization (https://www.mathworks.com/matlabcentral/fileexchange/7059-fft2-optimization), MATLAB Central File Exchange. Retrieved .
It slows down. Speed improvement is negative.
this problem is still present in MATLAB 7.1???