In this study, a scatter model is derived for parallel-beam, fan-beam, and cone-beam geometries in SPECT imaging. In the model, a photon is allowed to be scattered only once, and the probability of scatter for a given angle and energy is calculated using the Klein-Nishina formula. The detector is assumed to have perfect energy resolution. The scatter counts are computed for every projection array. From the scatter counts, the scatter line source response function and scatter-to-primary ratio are obtained. They agree well with those of Monte Carlo (MC) simulation including only single scattering, but deviate from those of full MC simulation including both single and multiple scattering. The deviation depends on the source depth within the medium. For a source depth of 6 cm, the difference of the scatter-to-primary ratio between the model and full MC simulation is less than 7%, while the difference becomes 27% for parallel-beam and 32% for cone-beam geometry at a source depth of 21.6 cm. Since scatter accounts for 20-40% of the total counts in most clinical studies, the scatter model yields a SPR accuracy ranged from 3% to 12%. The scatter model provides an efficient mean of characterizing scatter response with reasonable accuracy, and can be used in developing scatter compensation techniques in converging-beam SPECT.