Abstract: A seismic blind deconvolution method based on negative entropy was discussed.Blind deconvolution method does not take minimum-phase wavelet and Gauss white noise reflection coefficients as basic assumptions.Through utilizing Gaussian hybrid model,the strong (geologic targets) and weak (subtle reservoir) reflection information in reflection coefficient sequence were simulated;blind deconvolution objective function was defined in terms of non-Gauss criteria based on negative entropy;expectation maximization algorithm was applied to compute the optimal estimation of model parameters,finally non-Gauss reflection coefficient with high similarity to original reflection coefficient was obtained.Different phase wavelet and non-Gauss reflection coefficient model were selected to prove the method,and compared to Viener impulse deconvolution.The results show that the algorithm adapts non-minimum phase and non-Gauss system.The method was applied actual seismic data,the frequency bandwidth was effectively enlarged,and the resolution was improved.