In positron emission tomography (PET), a detector's response to scattered radiation may be different from its response to unscattered (true coincidence) photons. This difference should be accounted for during normalization and scatter correction. The disclosure shows that only a knowledge of the ratio of the scatter to trues efficiencies is necessary, however. A system and method are disclosed for measuring the scatter/trues detection efficiency ratio, as well as for applying this compensation during the scatter correction of PET emission data. PET detector efficiencies are measured in two steps, the first using a plane radiation source, and the second using a plane radiation source in combination with a scattering medium. A ratio of the scatter and trues detection efficiency is obtained from this data for each detector/crystal, and is applied as a correction factor to PET data obtained during medical imaging processes.