Abstract — This paper proposes a new method for directional-permeability assessment with nuclear-magnetic-resonance (NMR) measurements. Conventional techniques for permeability assessment from NMR measurements include empirical correlations such as SDR (Schlumberger-Doll-Research) and Coates models. However, carbonate rocks are known for lack of good correlations between pore-body-size and pore-throat-size, which makes it challenging and often unreliable to estimate permeability from NMR T2 (spinspin relaxation time) distribution in carbonate formations with complex pore structure. It also was proposed that conventional permeability models can be improved by incorporating an estimated pore-connectivity factor. However, none of the previously introduced techniques reflects the anisotropic characteristics of rock permeability. This paper introduces a new NMR-based directional-permeability model by incorporating a directional pore-connectivity factor into a conventional NMR-based permeability model. We introduce two approaches to quantify the directional pore-network connectivity of rock samples with pore-scale images. The first approach calculates directional pore connectivity in 3D pore-scale images with a topological technique. The second approach combines image analysis and electrical formation factor. The new NMR-based permeability model enables assessment of rock permeability in any desired direction. We successfully calibrated and tested the introduced NMR based permeability model on carbonate, sandstone, and sandpack samples with complex pore geometry or anisotropic permeability. The anisotropic permeability used for calibration and test purposes was obtained by the lattice Boltzmann method (LBM) simulations on microcomputed tomography (CT) images of rock samples. The comparison between the permeability estimates with our new AQ1 NMR model and conventional NMR models (e.g., SDR and Coates models) demonstrated that the NMR-based directional-permeability model significantly improves assessment of rock permeability, by reflecting rock’s anisotropic characteristics and minimizing calibration efforts. The outcomes of this research can significantly improve permeability assessment in complex carbonate reservoirs and anisotropic sandstone reservoirs, and can be extended further to organic-rich mudrock formations.