Coal matrix, the storage space for adsorbed gas, is highly related to gas recovery and gas outburst disasters. However, existing research predominantly relies on theoretical assumptions, lacking experimental validation. In this study, we employ X-ray computed tomography (CT) to obtain high-resolution (13.45–15.157 μm) internal structure images of various coal samples. The three-dimensional model of coal matrix elements is reconstructed using entropy threshold segmentation and watershed segmentation. A systematic quantification of the size, shape, and spatial arrangement of these matrix elements is then performed. The results reveal that coal matrix element equivalent diameters span 148.45–2800.35 μm, with volume-weighted average equivalent diameters ranging from 1320.82 to 1720.87 μm. Additionally, we explore particle size distribution (PSD) models to characterize size-quantity relationships and identify the Jaky model based on the logarithmic-exponential equation as the optimal PSD model. Observation of the matrix element models shows that large matrix elements typically exhibit flattened morphologies with rough surfaces, while smaller ones tend to be blocky with smooth surfaces. The dimensionless shape factor results suggest that, when scale effects are not considered, the shape of larger matrix elements is more conducive to matrix-fracture gas mass transfer. Based on this, a functional relationship between the matrix shape factor and size, considering changes in matrix morphology, is established. Moreover, we algorithmically determine that the average number of adjacent matrix elements per individual matrix element falls between 13 and 15. For matrix clusters with large volume fractions, the size ratio of the central matrix element to adjacent matrix elements primarily falls within the range of 0.4–0.6, approximately following a Weibull distribution. According to this feature, a method for building a matrix cluster model used for simulations is proposed.