Abstract: Overcomplete dictionaries can be constructed to sparsely represent targeted wave components or coherent interference based on their diverse morphologies for seismic signal analysis.This paper gives two applications of sparse representation based on overcomplete dictionaries, namely, ground-roll attenuation and interference suppression of the reservoir seismic response caused by a strong shield.For ground-roll attenuation, a low Q-factor wavelet function is constructed as the sparse representation dictionary for low-oscillatory body waves, and a high Q-factor wavelet function for high-oscillatory ground-roll.Thus, seismic data including body waves and ground-roll can be nonlinearly decomposed into low-oscillatory and high-oscillatory components.Testing on field data shows that this method can attenuate strong ground-roll noise in seismic records effectively, and preserve the waveform characteristics and spectral bandwidth of the effective signal well.For strong shield suppression, curvelet transform atoms are used to sparsely represent a strong shield, and two-dimensional stationary wavelet transform atoms are used to sparsely represent the reservoir seismic response.The results show that the method could attenuate the strong shield interference while preserving the effective reservoir seismic response.