Recent development of continuous recording systems used in industry, computational power, and very dense recording system makes seismic interferometry with ambient noise data an interesting tool to estimate structural information. To overcome the complexities of ambient noise, spectral normalization (e.g., whitening) is often used. Here I study different spectral normalization methods, such as crosscorrelation, deconvolution and crosscoherence, using numerical experiments and stationary-phase approximation. In the ideal situation of noise distribution, direct waves are accurately observed for all methods, but scattered waves are not. Especially, coda waves contain a lot of pseudo arrivals. Therefore, if we do not pay attention for the pseudo arrivals, we would use these waves for structural imaging and monitoring and create artifacts. Characterization of such waves using stationary-phase analyses sheds light on using and/or suppressing them.