Soil structure affects soil water and gas transport and soil biological activities, thus influence greenhouse gas (GHG) emission from soil. In this study, incubation experiment was conducted in two paddy soils (whitish paddy soil and red paddy soil) to study the effects of pore structures on GHG emissions under different treatments, i.e. non-puddling (NP), puddling (PD), and repacked after puddling (RP). Soil pore structure was determined by using the X-ray computed tomography and image analysis, and GHG emissions during rice growth period were measured by the static box method. The results showed that puddling significantly decreased soil macroporosity and pore connectivity, while NP and RP had more large pores with higher connectivity for both soils. PD promoted CH₄ emission from whitish paddy soil, which was 2.5 and 14.6 times of NP and RP, respectively. Correlation analysis showed CH₄ emission negatively correlated with macroporosity, indicating the increase of large porosity will reduce CH₄ emissions. NP had the highest CH₄ emission from red paddy soil, possibly due to the lowest porosity of ≤ 30 μm pores promoting CH₄ emission. NP significantly increased N₂O emissions from red paddy soil, which was negatively correlated with the 30-1 000 μm pores. The global warming potential (GWP) of RP for both soils was significantly lower than other treatments. In conclusion, this study confirms that the change in soil pore structure greatly affected GHG emissions and maintaining a porous soil structure can reduce CH₄ emission and global warming potential.