A method of processing accidental coincidence events (AC) and detector efficiency (DE) calibration data, which reduces the statistical noise in these measurements, and, consequently, reduces the noise in positron emission tomographic images using the technique, is described. The technique uses the fact that, in these measurements with N detectors in coincidence with N other detectors, N2 values of ACs or DEs are measured. However, these values are composed of only 2N components, which are either singles rates or individual DEs. The full set of data is used to implicitly solve for these values and the individual ACs or DEs recalculated with an improvement in statistical error equivalent to an N2/(2N + 1) increase in accumulated events for the case of a uniform distribution. This result was verified experimentally.