A new incremental approach to the computation of vertical excitation energies is presented. The method works based on the definition of a dominant occupied orbital where the excitation takes place (natural transition orbital) and by localizing the remaining occupied space. The use of a reduced two-body expansion leads to a linear number of terms to be computed. A series of benchmark calculations have been carried out on small to medium sized photoactive systems. The results compare well to the full calculations, with maximum deviations of 0.3 eV, and an average absolute deviation of about 0.08 eV. In addition, a energy decomposition analysis is made on the basis of orbital distances to the chromophore region. First results indicate that orbitals beyond a relatively small radius can be safely neglected, leading to further drastic savings in the calculations.