Size segregation by self-diffusion during gas-phase flow of coarse binary granular mixtures is quantified using a novel dual energy gamma-ray photon technique which is capable of producing solids fraction profiles for each of the individual components of a binary mixture in addition to the voidage profiles. Spatial and temporal data of solids fractions in a binary mixture are analysed using methodology based on statistical mechanics principles which lead to the definition of 'micro-turbulence' during flow in terms of the local fluctuating volume fractions of individual solid components. The self-diffusion velocities and the diffusion coefficients of coarse and fine particles in the mixture are calculated based on an analysis of local random drift during the flow of a particle assembly. Th:se calculations are also compared with theoretical predictions of self-diffusion coefficients based on the kinetic theory which assumes random collisions between inelastic particles.