Abstract: As a basic method to quantify the scale effect of porous media materials, representative elementary volume (REV) is an effective means to construct the macro-meso structure of coal. In order to improve the accuracy of REV results of coal meso-structure, a quantitative characterization method of coal meso-structure was proposed based on the study of pore-fracture structure characteristics at meso-scale of coal samples. Based on the binary images of pore-fracture at 423 scales at different positions and different fracture directions of coal samples, the geometric characteristics of coal meso-structure were analyzed from three aspects: pore-fracture distribution density, geometric shape and orientation. Based on the sampling method considering the heterogeneity and anisotropy of coal, the spatial effect and directional effect of REV size are analyzed. Finally, the reasonable microscopic REV size of coal sample is determined by the coefficient of variation method. The results show that: ① A method for accurate identification and quantitative analysis of pore fracture structure is proposed for SEM images of coal samples with uneven gray level, blurred boundary and low contrast. ② A quantitative characterization system of meso-structure reflecting the anisotropy of coal samples was established, including porosity, overall shape factor, fractal dimension and directional distribution coefficient. ③ Compared with the parallel bedding, the micro-fractures in the vertical bedding direction are more developed, the porosity is 2 times that of the parallel bedding, the contribution rate of large aperture pores is 85%, and it has strong spatial filling ability. The overall morphology of pores and fractures is poor, but the edge morphology is smooth and the directional distribution tends to be simpler. ④ Based on the acceptable value of the coefficient of variation of 10%, the size of the coal meso-structure characterization unit is determined to be 200 μm × 200 μm.