Tight sandstone gas zones displayed such characteristics as low porosity and low permeability. At present, it was a challenge for log analysts to gain reservoir absolute permeability precisely in such zones because of the poor correlation between formation porosity and permeability parameters. Classical SDR model and Tim-Coates model from nuclear magnetic resonance (NMR) log data also losed their advantage. In this condition, 21 core samples were chosen to run mercury-injection and NMR log experiment simultaneously, whose porosity ranged from 2.7% to 16.1% and permeability ranged from 0.019 × 10−3μm2 to 131.0 × 10−3μm2. With the analysis of mercury-injection data, correlations were established between rock absolute permeability and the pore structure parameters, such as the average pore throat radius. The inflexion parameter of capillary pressure curves (the Swanson Parameter) was also associated with the rock absolute permeability. By using these relationships, the rock permeability could be calculated from mercury-injection data. Considering the problem that mercury-injection capillary pressure data was limited by their quantity because of the expensive test cost and the environmental concern, based on the fact that NMR T2 distribution and mercury-injection capillary pressure curves all reflected reservoir pore structure, two novel models were established to gain the average pore throat radius and the Swanson parameter from nuclear magnetic resonance T2 distribution. So, these parameters could be calculated from NMR imaging log data and then rock absolute permeability could be gained respectively by combining NMR total porosity. It had demonstrated that these clues could be expandeded to reservoir conditions, and then reservoir absolute permeability could be calculated along the intervals with NMR imaging measurement. At last, an in-situ example from southwest oil and gas field of China was exhibited of calculating reservoir absolute permeability consecutively in tight sandstone gas zones by using these two models and the validity of these models were compared. The Swanson parameter model was suggested in calculating reservoir absolute permeability in this paper. The result shown that absolute permeability calculated by using the Swanson parameter model matched much better with core data than the average pore throat radius model.These two models were also effective and accurate in detecting the thin sand with high porosity and high permeability because a core sample used in these models came from these zones.