Spectral decomposition has been widely used in seismic interpretation. Many methods are proposed, such as S transform and pseudo Wigner-Ville distribution for spectral decomposition. These two methods are insufficient in the time resolution or clarity. A method which combines matching pursuit and pseudo Wigner-Ville distribution is applied to enhance the time resolution and eliminate cross term interferences. This method decomposes the input signal into a series of wavelets belongs to a preset dictionary, then analyze time-frequency character for each wavelets using pseudo Wigner-Ville distribution and stack the results together. It keeps good time resolution in pseudo Wigner-Ville distribution. Simultaneously, cross term interference is eliminated because matching pursuit transforms the input signal into multiple independent wavelet signals. This method is applied on the synthetic trace and real seismic section. Its result which has higher resolution and elimination of cross terms interference can be used to indicate the existence of hydrocarbon. Matching pursuit-based pseudo Wigner-Ville distribution has a better performance in spectral decomposition, though it is more time consuming. © 2017 SEG.