Penguin populations are potentially sensitive indicators of ecological change in Antarctic and sub-Antarctic marine ecosystems. Aerial photographic surveys provide the most robust method for estimating breeding population size, particularly for large colonies. Obtaining population estimates from aerial photographs is laborious and usually carried out by manually counting individual birds on highly magnified prints. I derived population estimates using computer-based image analysis of digitally scanned color aerial photographs of macaroni penguin (Eudyptes chrysolophus) colonies at Bird Island, South Georgia. I compared automated image analysis with manual counts from the photographic prints and conventional ground counts, highlighting assumptions that contributed to differences in population estimates. The automated image-analysis routines produced estimates that were highly correlated with ground counts, indicating that the technique could be reliably used for large-scale macaroni penguin population surveys.