Abstract: Formation lithology hydrocarbon reservoirs are mainly developed in Y1 well area of Junggar basin. But the reservoir recognition and prediction are quite difficult because of deeply buried interests (more than 5,000-6,000 m), thin reservoirs (8-10 m), small difference between reservoir impedance and surrounding rock impedance, weak reflection energy and interference of coal seams. Aiming at these problems, the inverse Q filtering method was used in the 3D seismic data to improve the resolution. Firstly, phase correction was carried out on the selected cross-well seismic profiles and compared to the synthetic seismogram of logging data to obtain optimum phase correction operator. Then, the amplitude was compensated and the compensation operator adjusted through spectrum analysis to ensure that the Q filtering will not damage the relative variation relationship of all effective frequencies during amplitude compensation. The optimum Q modeling was achieved through repeating the abovementioned process. The model has characteristics of time-variation and space-variation. The application of it in this area shows that the resolution of seismic profiles is obviously enhanced, the recognition ability of formation lithology traps and reservoir prediction accuracy improved, and the updip wedge-out and sand body distribution range identified.