Abstract: The tight sandstone reservoirs of the braided river delta facies in the Xujiahe Formation of the Sichuan Basin represent a critical target for the reserve growth and production addition of hydrocarbons. In the TJ area, the Xujiahe Formation exhibits deep burial depths, complex lithologies, unconfirmed petrophysical patterns of dominant lithofacies, and rapid lateral variations in reservoirs, which jointly limit the reservoir prediction accuracy. This study aims to address the challenges in reservoir prediction. First, considering the impacts of multi-mineral components, a generalized Xu-White petrophysical model was developed to accurately determine the factors for lithologic and reservoir identification. Then, using a reflectivity objective function with time-offset dual constraints developed in this study, the amplitude versus offset (AVO) reflectivity extraction method was improved, contributing to the acquisition of accurate ternary reflectivity volumes. Accordingly, high-resolution prestack elastic parameter volumes were determined through Bayesian inversion. Finally, the distribution of sandstone reservoirs was predicted thoroughly by combining factors for lithologic and reservoir identification, as well as high-resolution prestack inversion results. The improved AVO reflectivity inversion accurately predicted the distribution of sandstones and reservoirs in the TJ area, delivering satisfactory performance in predicting highly heterogeneous tight sandstone reservoirs. This study provides a novel approach to prestack prediction of tight sandstone reservoirs.