The trend of increasing pattern data volume, together with decreasing critical dimensions, causes correction of proximity effect to be continually more time consuming and expensive. A tradeoff exists between speed and accuracy. A given correction method may be judged efficient if the direct computational overhead needed to achieve a predetermined level of accuracy is small. Some basic principles relating to efficiency are derived. An optimal correction algorithm follows naturally from the analysis. Forward and backward scattering have greatly differing ranges, and correspondingly differing computational requirements. The two processes are easily separated analytically. Dose correction is used to offset backscattering, while edge displacement is used to offset forward scattering. The only direct computational overhead is the local pattern density. All other necessary parameters are precomputed, and accessed by memory lookup.