Plant leaves have heterogeneous structures composed of spatially variable distribution of liquid, solid, and gaseous matter. Such contents and distribution characteristics correlate with the leaf vigor and phylogenic traits. Recently, terahertz (THz) techniques have been proved to access leaf water content and spatial heterogeneity distribution information, but the solid matter content and gas network information were usually ignored, even though they also affect the THz dielectric function of the leaf.A particle swarm optimization algorithm is employed for a one-off quantitative assay of spatial variability distribution of the leaf compositions from THz data, based on an extended Landau-Lifshitz-Looyenga model, and experimentally verified using leaves. A good agreement is demonstrated for water and solid matter contents between the THz-based method and the gravimetric analysis. In particular, the THz-based method shows good sensitivity to fine-grained differences of leaf growth and development stages. Furthermore, such subtle features as damages and wounds in leaf could be discovered through THz detection and comparison regarding spatial heterogeneity of component contents.This THz imaging method provides quantitative assay of the leaf constituent contents with the spatial distribution feature, which has the potential for applications in crop disease diagnosis and farmland cultivation management.