Interference data obtained by a spatial heterodyne spectrometer (SHS) is subject to various error factors and suffers from complex phase distortion. Traditional phase correction methods, such as the Amplitude, Merzt, and Forman methods, extract phase distortion in the spectral domain and correct it, which cannot effectively correct spatial phase distortion. Through theoretical derivation and numerical simulation, the spatial phase distortion is firstly determined and corrected in the interference domain. The frequency-dependent phase distortion is then extracted in the spectral domain and corrected. This novel phase distortion correction method named the phase decomposition method was applied to the in-orbit interference data of Greenhouse gases Monitoring Instrument-II (GMI-II). Compared with traditional phase correction methods, the root-mean-square error of the spectrum corrected using the phase decomposition method is reduced by 81.37%.