Abstract: The structural-lithologic reservoir has the characteristics of thin thickness of sand body, rapid lateral change of reservoir and complex fault boundary. Therefore, the seismic prediction and boundary description of this oil reservoirs are challenging with low accuracy. In order to improve the prediction accuracy of thin sand reservoir, a seismic data processing method is innovatively proposed in this study by using steerable pyramid decomposition technology, which mainly includes the creation of pyramid structure (image decomposition) and image reconstruction after decomposition. The multi-scale decomposition and combination of seismic data allow to extract the geological body information in different scales and directions. Then the combination of seismic data of different scale is selected based on requirements of geological body description, so as to highlight and describe the characteristics of geological body from macro and micro perspectives. Finally, the seismic attributes and seismic inversion study are carried out on seismic data processed by this method. The reservoir prediction results are more in line with the rules of geological body when seismic data are selected with proper scale, avoiding the limitations of describing macro geological rules by using small-scale data and the uncertainty of characterizing micro heterogeneity of sand body by using largescale data.