The Fourier transform can express a function in the time domain as a function in the frequency domain, which means every discrete 1D-signal is actually equal to the sum of its discrete frequency components(0Hz, 1Hz...Fs/2 Hz) with the components' phases respectively, so we can use the frequency components to reconstruct the original signal. For 2D-signal(complex number), its the same idea, just the frequency components become(0Hz, 1Hz...Fs Hz). We express the components in the form of spinning circles, so that the reconstruction of the signal becomes drawn by connected circles.