Happily, Fourier sums are exactly invertible: given the output, the input can be quickly found. Because signals can be transformed to the frequency domain, manipulated there, and then returned to the time domain, convolution and correlation can be done faster. Time derivatives can also be computed with more accuracy in the frequency domain than in the time domain. Signals can be shifted a fraction of the time sample, and they can be shifted back again exactly. In this chapter we will see how many operations we associate with the time domain can often be done better in the frequency domain. We will also examine some two-dimensional Fourier transforms.

- FT AS AN INVERTIBLE MATRIX
- INVERTIBLE SLOW FT PROGRAM
- SETTING UP THE FAST FOURIER TRANSFORM
- TWO-DIMENSIONAL FT
- References

10/31/1997