The characterization of fracture morphology in steel fiber reinforced concrete (SFRC) remains an area of limited study. This research investigates the influence of heterogeneity on fracture surfaces in SFRC across various surface sizes, sampling intervals (δ), and directions. Utilizing point-cloud matching technique, fracture morphologies are characterized to analyze damages in the heterogeneous fracture process zone. Four roughness parameters, i.e., root mean square deviation (Sq), arithmetic average height (Sa) distribution, surface area to projective area (RS) and root mean square of directional derivatives (Z2–3D), are used to quantify morphology. Results reveal that as surface size arises, Sq and Sa increase with their ratio (Sq/Sa) approaching 1.25, while the ratios of Sq and Sa to size length decrease following a double-exponential decay model, indicating a size scale dependence. Meantime, roughness anisotropy, characterized by coefficient of variations in directional Z2–3D values, declines. RS and Z2–3D exhibit higher sensitivity to textural features than surface size, showing a δ-dependence. As δ rises, RS and Z2–3D sharply decrease, whereas Sq, Sa and Sq/Sa stabilize. This analysis highlights the significance of Z2–3D in providing insights into fracture morphological features and heterogeneity of SFRC.