Seismic petrophysical inversion is an effective method for reservoir physical property evaluation. Direct prediction of reservoir parameters from seismic data has lower uncertainty and higher accuracy than estimation of reservoir parameters from seismic elastic parameters. However, at present, there is little discussion on the establishment of initial model in direct reservoir parameter inversion. A reasonable initial model can not only improve the accuracy of inversion results but also reduce the calculation cost of inversion process. To solve this problem, this paper proposes a seismic reservoir characterization method based on pre-stack and post-stack joint inversion, which combines post-stack impedance inversion and statistical rock physical model to provide a reliable initial model for pre-stack seismic rock physical inversion, and makes full use of the high signal-to-noise ratio of post-stack seismic data and the high resolution of pre-stack seismic data to improve the stability and accuracy of reservoir parameter inversion. Firstly, the critical porosity model is calibrated based on the existing logging data, and the reservoir parametric reflection coefficient formula is constructed based on Zoeppritz reflection coefficient equation, which establishes the direct relationship between seismic data and reservoir physical properties. Then the P-wave impedance is obtained by post-stack inversion, and the initial model of reservoir physical parameter inversion is obtained by using the statistical petrophysical model obtained from logging data. Finally, based on Bayesian framework and Cauchy prior constraints, the inversion of physical property parameters such as porosity, shale content and water saturation from pre-stack seismic data is realized. The synthetic tests show that the superior anti-noise performance of post-stack impedance can provide a reliable initial model for reservoir parameter prediction, and can significantly improve the accuracy of physical property inversion. The field data test verifies the advantages of this method in improving inversion accuracy and enhancing lateral continuity in direct estimation of reservoir physical properties.