This study focuses on the determination of the fraction of vegetation cover (fCover) based on the inversion of a four-parameter model combining the reflectances in the Red (R) and Near Infrared (NIR) domains. This model is semi-empirical since it is based on radiative transfer modeling, but requires parameter calibration related to SAIL simulations (Verhoef). As shown by Yoshioka et al., if the multiple soil/vegetation interactions are smaller than the first order ones then the fCover isolines can be approximated by straight lines in the (Red, Near Infrared) plane. Each isoline is completely defined by its slope α and its intersection point γ (with the soil line), that have been related to fCover using SAIL simulations and optimization either by the Simplex (local optimization) or the SCE–UA (global optimization) algorithms. The results are compared to classical vegetation indices for both simulated and actual data. The method shows an improvement in most of cases. Moreover, when using the SCE–UA algorithm, our approach proves its robustness relative to high noise level.